ICT TODAY

THE OFFICIAL TRADE JOURNAL OF BICSI

May/June 2016
Volume 37, Number 3

+ Today’s DCIM Solutions

PLUS

+ Optical Fiber to the Classroom

+ Raising Data Center Temperatures
 May/June 2016/Volume 37, Issue 3 CONTENTS
05 FROM THE PRESIDENT
Growing BICSI Around the Globe 28 Taking the Network to 40/100 Gigabit
Ethernet: Keys to installing a high-performance
By Brian Ensign, RCDD, NTS, OSP, RTPM, CSI optical fiber structured cabling infrastructure that
is essential to a successful migration to 40/100 Gb

06 COVER STORY
Designing Wi-Fi Networks for High
Ethernet. By Josh Taylor

Capacity: Wi-Fi networks for enterprises and

public venues can be implemented in one of 36 Optical Fiber to the Classroom: Significant
differences between two popular topologies for
two ways: careful design upfront, or constant deploying optical fiber to the classroom can
troubleshooting and adjustments throughout impact cost of deployment, total cost of owner-
the life of the network. By Jussi Kiviniemi ship and achievable bandwidth. By Aaron Hesse,
RCDD, PE

14 Data Center Infrastructure Management

Solutions—How to Hit a Moving Target:
Learn how scalability and modular capa- 44 DisplayPort and the Evolution of Desktop
Video Connectivity Solutions: A review of the
bilities allow businesses to choose DCIM latest developments in DisplayPort, the digital
bundles that meet their immediate needs display interface now found on the majority of
while allowing for future growth. By Bob Potter new convertible, laptop, desktop and workstation
computers. By Joseph D. Cornwall, CTS-D, CTS-I

18 Speaker Placement and Wiring

Guidelines for Public Address/Paging
System/AV: Regardless of the size or 48 The Unexpected Impact of Raising Data
Center Temperatures: Data center operators
specific use of PA/paging/AV systems, should consider these critical factors before making
speaker spacing and location are a key the decision to raise temperatures in their IT space.
part of ensuring that the system provides the By Wendy Torell, Kevin Brown and Victor Avelar
required degree of coverage. By Robert B.
Hertling Jr., RCDD, OSP

SUBMISSION POLICY
ICT TODAY is published bimonthly in January/February, March/April, May/June, July/August, September/October, and November/December by BICSI, Inc., and is mailed Standard A to BICSI members,
RCDDs, RITPs, RTPMs, DCDCs, BICSI Installers and Technicians and ESS, NTS, OSP and WD credential holders. ICT TODAY subscription is included in BICSI members’ annual dues and is available
to others through a purchased yearly subscription.
ICT TODAY welcomes and encourages submissions and suggestions from its readers. Articles of a technical, vendor-neutral nature are gladly accepted for publication with approval from the Editorial
Review Board. However, BICSI, Inc., reserves the right to edit and alter such material for space or other considerations and to publish or otherwise use such material. The articles, opinions and ideas expressed
herein are the sole responsibility of the contributing authors and do not necessarily reflect the opinion of BICSI, its members or its staff. BICSI is not liable in any way, manner or form for the articles, opinions
and ideas, and readers are urged to exercise professional caution in undertaking any of the recommendations or suggestions made by authors. No part of this publication may be reproduced in any form or by any
means, electronic or mechanical, without permission from BICSI, Inc.
ADVERTISING: Advertising rates and information are provided upon request. Contact the BICSI Sales Department for information at +1 813.979.1991 or 800.242.7405 (U.S. and Canada toll-free) or
sales@bicsi.org. Publication of advertising should not be deemed as endorsement by BICSI, Inc. BICSI reserves the right in its sole and absolute discretion to reject any advertisement at any time by any party.
POSTMASTER: Send change of address notices to BICSI, Customer Care, 8610 Hidden River Pkwy, Tampa, FL 33637-1000; Phone: +1 813.979.1991 or 800.242.7405 (U.S. and Canada toll-free)
© Copyright BICSI, 2016. All rights reserved. BICSI and RCDD are registered trademarks of BICSI, Inc.

May/June 2016 t 3
 ICT TODAY
THE OFFICIAL TRADE JOURNAL OF BICSI

2016 BICSI BOARD OF DIRECTORS ADVERTISER’S INDEX

President Brian Ensign, RCDD, NTS, OSP, RTPM, CSI
AFL............................................... Back Cover
President-Elect Jeffrey Beavers, RCDD, OSP, CFHP aflglobal.com
Secretary Robert “Bob” S. Erickson, RCDD, NTS, OSP, RTPM, WD
Corning Cable ................ Inside Front Cover
Treasurer Christy A. Miller, RCDD, DCDC, RTPM, CT
corning.com
Canadian Region Director José Mozota, Ph.D.

U.S. North-Central Region Director Chris Scharrer, RCDD, NTS, OSP Dura-Line......................................................13
duraline.com
U.S. Northeast Region Director Matthew Odell, RCDD

U.S. South-Central Region Director Todd W. Taylor, RCDD, NTS, OSP Hitachi..........................................................17
U.S. Southeast Region Director Charles “Chuck” Wilson, RCDD, NTS, OSP hca.hitachi-cable.com

U.S. Western Region Director Larry Gillen, RCDD, ESS, OSP, CTS
Hyperline.......................... Inside Back Cover
Executive Director & Chief Executive Officer John D. Clark Jr., CAE hyperline.com

EDITORIAL REVIEW BOARD ICC................................................................11

icc.com
Chris Scharrer, RCDD, NTS, OSP, WD

Jonathan L. Jew OFS................................................................41

ofsoptics.com
F. Patrick Mahoney, RCDD, CDT

PUBLISHER
BICSI, Inc. 8610 Hidden River Pkwy., Tampa, FL 33637-1000
Phone: +1 813.979.1991 Web: www.bicsi.org
CONTRIBUTE TO
ICT TODAY
EDITOR ICT Today is BICSI’s premier publication
Steve Cardone, icttoday@bicsi.org that aims to provide authoritative, vendor-
neutral coverage and insight on next
generation and emerging technologies,
PUBLICATION STAFF standards, trends and applications in the
global ICT community. Consider sharing
Wendy Hummel, Creative
your industry knowledge and expertise by
Amy Morrison, Content Editor becoming a contributing writer to this
informative publication.
Clarke Hammersley, Technical Editor

Jeff Giarrizzo, Technical Editor Contact icttoday@bicsi.org if you are

interested in submitting an article.
Allen Dean, Technical Editor

ADVERTISING SALES
+1 813.979.1991 or sales@bicsi.org

4 u ICT TODAY
 FROM THE PRESIDENT, BRIAN ENSIGN, RCDD, NTS, OSP, RTPM, CSI

The primary application for 40GBASE-T is very specific and the market is clearly
Growing BICSI Around the Globe
defined—data center interconnection from servers to edge switches.

available common speed between m (328 ft)

In my last column, four-connector
I outlined my reference
in Mumbai.important
I look forwardfeature of a category is
to meeting
two different connected ports,plans and channel implementation
for the upcoming term to support to 30 with m members there and doingcategory
its frequency. Since all I can6A and
Energy Efficient Ethernet (EEE), (98 ft) with
membership, strategic two connectors.
planning and The category 7 technologies are both
to advance MemberA Pride!
which affords significant savings via designated frequency for category being upgraded to 2000 MHz, the
volunteerism within BICSI. I am happy The rest of the first half of the year
power management by enabling a 8 twisted pair cabling media is respective new component sets
to report progress is being made on includes conference
for ISO/IECattendance
category 8.1inand 8.2
link sleep mode. 2000 megahertz (MHz). The two-
Most of the discussion allforfronts!
the connector channel proposed byCanada, TIA Rome
will work andtogether
London, to asenable
well asmore
new category 8 cabling is not Task forces are in place to address
will use two backward-compatible budget and business meetings
channel implementation at BICSI
options.
focused on the traditional LANand make recommendations
RJ45 connectors, forandnext- the cable Headquarters
will TIA
in uses the
Tampa. category
It will be a very name
model. The objective is set generation
for two- development
likely be anand outreach,
unshielded to identify components,
busy, but very rewarding time for me.
twisted-pair links and
connector channels up to 30 withmeters (UTP) core
updates expected bywith
June. overall channels,
We shielding These outreach programs will by whereas the ISO/IEC
(m [98 feet (ft)]). Accordingly, short- that will appear
have kicked off Credentialing 2020 to similar tono means standard
bewith designates channels and
take any focus away from
reach data center interconnection screened and category 6 (i.e., F/UTP). links as “classes,” which are built
a period of fact-finding auditing. our efforts from
in all the
interfaces will be the first applica- These shielded cables will be IEC componentNorth
Global and “categories.”
I am pleased with the level of Member American For markets
example,we serve
Class today.
E channels are
tions supported by 40GBASE-T. necessary for structured cabling, A
Pride I am seeing through my meetings, Remember: a member
normally builtis a member;
from category 6A
In the development of category however very short channels might
social
8 cabling, the current upper two media and global travel.
still be possible with UTP cable. their needs components.
are our needs; ISO/IEC
and will
there therefore
This is a busy time at BICSI.terminations
It’s budget more define
are no borders two new
within BICSI.channel classes—
levels of cabling performance, Making shield
season, and we are preparing for the Class-I and
Finally, I should mentionClass-II—tothat thebe call
built
extended category 6A and category efficient and robust is an immediate,
from category 8.1 and category 8.2
7A technologies, have bothnext fiscal yearserious
survived startingand Julyworthwhile
1. What’s challenge for nominations is out through May 31 for
components respectively.
the past year of preliminarymore, the global to conference
category 8 componentschedule designers. the upcoming While Board category
of Directors elections
7A (Class FA)
discussions. The category 6A-based
is heavy. In fact,The good
I am newsthis
writing is that
column for category
for the following positions:
suppliers might be expected to be
cabling technology offers fully 6A and up, shielded F/UTP cables
from Dubai, UAE, where I am attending u Secretary
already well on their way to making
backwards compatible connectivity are inherently smaller than UTP
the BICSI Middle East and Africa (MEA) u Canadian
category Region8.1Director
components, that is
for short-reach applications. cables. Furthermore, with the
District Conference & Exhibition. We made u Globalnot Regionnecessarily
Director the case. Category 7
(new)
The category 7A-based cabling RJ45 remaining to be one of the
two key announcements here to support has been available
U.S. South-Central Region Director since 1996 and
technology offers higher bandwidth smallest connectors used in theudata
category 7A since 2006, but they
the goals set incenter
LAN connectivity for supporting our strategic plan: to40GBASE-T
environment, u U.S. Southeast Region Director
have not been standardized beyond
continue our global
extended reach to future multimedia offers reach
quadruple with a density u U.S. Western
focus
the Region Director
1200 MHz. Nevertheless, most
applications. This article explores
on Dubai and the overMEA 10GBASE-T—pathways
District. First, we and manufacturers are aware of category
the potential cable and connector
announced that faceplates will getafour
BICSI received licensetimes more I am delighted with the level of
7 and 7A international standards
technologies under consideration
from thefor Dubaicapacity
Association fromCentre
the same to connector.questions and
and comments
have adopted I have received
portions of the
the category 8 standards, their The International Standards from potential candidates,
technology into which
their shows
products.
officially operate and be recognized as
impact on the information tech- Organization (ISO)/International me the interestBoth is high for these
category 8.1positions,
and category
an association here. Second, the open-
nology systems (ITS) industry and Electrotechnical Commission and for a good 8.2 cables will use the
reason: it’s a great samewaygauge
ing of the BICSI(IEC)
what the future holds for twisted- Dubai office is is
standard targeted
spawning two wire. The key distinction setting
to volunteer!
pair cabling beyond category for 8.
October 2016. Theseof
versions developments
category 8—category category 8.2 apart is the use of
will provide the8.1 local Things are moving quickly, and I look
andsupport
category needed
8.2. Derived individually-shielded twisted-pair
What Will it Look Like? to expand our from reachcategory
to our members
6A, category 8.1 is forward to my next update. Until then,
cables (F/FTP), traditionally used for
The TIA standard for category essentially equivalent to the TIA keep showing that Member
7 and 7A.Pride!
in Dubai and the MEA District. category Like TIA category
8 is being developed as a seamless category 8 standard.
After Dubai, our next stop is India, On the other 8, the designated connector for
upgrade from category 6A. whereThe we will hold hand, categoryin8.2
meetings is derived
Delhi ISO/IEC category 8.1 is the RJ45.
application provides an important from fully-shielded category 7A Category 8.2 will use upgraded non-
and attend the BICSI India Conference
relaxation from the traditional 100 technology. Remember, the most RJ45 category 7A-type connectors.

May/June 2016 t 5
6 u ICT TODAY
By Jussi Kiviniemi

We have all experienced Wi-Fi networks that failed

to meet our expectations, causing us minor
inconveniences. A Skype call drops at the airport.
A Facebook update fails at a crowded stadium.
But the consequences can be serious, too; imagine
hospital equipment failing to communicate wirelessly.

Why is it that Wi-Fi often works Coverage Is King

very well at home but fails in many
enterprises and public venues? The
main reason is the demand for
increased capacity combined with poor
Wi-Fi network design. Because only
a few people are using the network at
home, the Wi-Fi network does not need
specific design considerations. But
when it comes to public places,
corporate offices and warehouses,
Wi-Fi networks get more crowded as
more devices join. There is also more
variation between mobile devices
including laptops, tablets, smartphones,
Wi-Fi enabled hospital equipment,
forklift computers, Wi-Fi phones and
Wi-Fi-based location tracking tags. The
network needs to be designed to support
all of these devices.
For an IT administrator, Wi-Fi can
be implemented in two ways:
u Either more attention is paid to
designing the network; or
u The network is built quickly,
but constant troubleshooting
and adjustments are required.
A high-capacity Wi-Fi network
is one that allows more data to be
transmitted wirelessly than a Wi-Fi
network built just for coverage. Simply
adding Wi-Fi access points (APs) does
not necessarily mean high capacity—
but it can, when combined with smart
frequency reuse and other tips presented
in this article.
Whether or not it is designed
specifically for high capacity, the Wi-Fi
network must be designed for sufficient
coverage. The level of coverage is a
factor of the strength of the signal trans-
mitted between the AP and the mobile
device. If the signal is too weak, the
connection will be unreliable or will
break down completely.
Oversimplified, the closer the
mobile device is to the AP, the
higher the signal strength. But it is
not just distance that weakens the
signal; everything that comes between
the AP and the mobile device—building
walls, people, furniture—attenuates
Wi-Fi signals.

May/June 2016 t 7

It is also imperative to note that signal strength is a
two-way street. It is not enough that the mobile device
can hear the AP; the AP needs to hear the mobile device,
as well. Therefore, even if the AP radio power levels are
increased to the maximum and the mobile device can
hear the AP very well, the connection still might not work
due to a failure from the mobile device back to the AP.
In Wi-Fi networks, distributed antenna system (DAS)
solutions are not commonly used. Instead, the antenna
is located close to or, more typically, built into the AP
(Figure 1). That, however, does not mean the coverage
pattern is circular, as it is affected by walls, elevator shafts,
and other objects. Because of this, specific Wi-Fi planning
tools are used to design coverage that meets the user
requirements.
In areas that are challenging for radio network design,
such as warehouses, the APs may be mounted 15 meters
(m [50 feet (ft)]) up in the ceiling, while the client devices
are on the floor level. In cases like these, directional
8 u ICT TODAY
FIGURE 1: Example of simple Wi-Fi coverage planning.
Courtesy of the National Telecommunications
and Information Administration
FIGURE 2: 5 GHz channels in the U.S.
antennas are required to direct the signal toward the
client devices. Antennas must be tilted horizontally, as
well as vertically. 3D planning tools help in figuring out
how to align the antennas.
In the end, what is sufficient coverage? For a data-
only network, it may suffice to provide a signal strength
of -75 decibel milliwatts (dBm) or better. However,
-75 dBm signal strength may not be enough to achieve
crystal-clear voice call quality. Values ranging between
-65 to -67 dBm are often referred to as industry standards
for deploying high-quality voice networks.
Coverage refers to the signal strength received from
the strongest of the audible Wi-Fi APs. However, to
ensure Wi-Fi connectivity when roaming from one AP
to another, sufficient signal strength from the second
strongest AP is also required. A typical design guideline
would be -67 dBm signal strength everywhere for the
strongest AP, and -75 dBm signal strength everywhere
for the second strongest AP.
 g/
lo
.com/b
hitewifi
.greatw
/www
: http:/
g
Fi Blo
ite Wi-
at Wh
© Gre
FIGURE 3: APs placed in hallways and all 2.4 GHz radios
enabled, causing co-channel interference (gray).
Minimizing Channel Overlap
Wi-Fi networks operate on two frequencies: 2.4
gigahertz (GHz) and 5 GHz. Both frequencies have a
number of available, interference-free channels. 2.4 GHz
has three interference-free channels, and 5 GHz has more
than 20 in the U.S.—but all may not be usable (e.g.,
because of interference with radar systems [Figure 2]).
Most Wi-Fi APs have two radios that operate
simultaneously: one operating on 2.4 GHz frequency
and one on 5 GHz (some of the newer APs are capable of
configuration for dual 5 GHz operation).
Mobile Wi-Fi devices typically support both 2.4 GHz
and 5 GHz frequencies but operate on just one of them
at a time. Many older devices, as well as Internet of
Things (IoT) devices, support just 2.4 GHz, preventing
many networks from disabling the 2.4 GHz frequency
space altogether.
Because 2.4 GHz frequency space needs to be
supported, and it only has three non-overlapping
channels to work with, channel planning becomes
a bit of a challenge to achieve high network performance.
In 5 GHz frequency space, the 20+ channels are often
reduced down to a few due to radar, other networks,
non-Wi-Fi interference, supporting legacy devices and
especially the use of wider channels (one AP may take
the channel space of two or even four APs).
The network should be designed to
minimize the number of overlapping channels.
Ideally, this means that in every location, only
one AP should be operating on channel number
1, for example. This sounds easy, but when
there are just three channels to work with, plus
neighboring networks, plus interference from non-
Wi-Fi devices, it becomes challenging, especially in
a multi-floor environment with thick walls and open
spaces close to one another.
The key to mitigating co-channel interference
is careful channel and transmit power allocation in
combination with coverage design. Designing for
coverage—which is relatively easy using modern tools—
will be followed by fine-tuning the channels, as well as
the locations of the AP and the transmission powers to
minimize channel overlap.
A rule of thumb for avoiding channel overlap in
many indoor environments is the placement of APs in
rooms instead of hallways. This way, the wall absorption
reduces the AP signal fingerprint, leaving less chance for
channel overlap (Figure 3).
On the 5 GHz frequency, the IEEE® 802.11n standard
introduced 40 megahertz (MHz)-wide channels as
an option for the traditional 20 MHz-wide channels.
802.11ac introduced 80 MHz-wide and even 160 MHz-
wide channels. The idea is simple: double the frequency,
double the network throughput. However, if not used
carefully, wider channels may lead to significant channel
overlap, even on 5 GHz, and thus reduced, rather than
increased, throughput. Especially in high-density network
layouts, 20 MHz is still often recommended. 40 MHz is
becoming more of a common choice, while 80 MHz is
still being used selectively. With 2.4 GHz on enterprise
wireless LANs (WLANs), 40 MHz wide channels should
never be used.
There are two limiting factors with regards to
available Wi-Fi channels on the 5GHz frequency space:
client device support and radar. Enter DFS (Dynamic
Frequency Selection), a mechanism implemented for
sharing the spectrum between Wi-Fi and radar. When an
AP is operating on any channel between 52 and 1404, it
is mandatory for the AP to monitor radar activity. If radar
is detected, the AP must cease from transmitting data and
switch to a different channel. When it comes to client
devices, the so-called DFS channels can be completely
May/June 2016 t 9
 FIGURE 4: Wi-Fi signals appear as curvy shapes in a spectrum analyzer.

unsupported by the device manufacturer, especially with
older 5 GHz client devices. DFS support on the APs, as
well as the client device channel limitations, calls for extra
caution with Wi-Fi channel assignment, especially with
outdoor Wi-Fi deployments.
Since one AP typically includes both 2.4 GHz and 5
GHz radios, and the 5 GHz frequency space has multiple
times more interference-free channels, it often becomes
impossible to find a satisfactory channel plan for 2.4 GHz
without turning off some of the 2.4 GHz radios. It may
sound counterintuitive, but disabling some of the 2.4 GHz
radios to minimize overlap actually increases capacity.
Penalty for Interference
A question often heard is “Why does our wireless
network become unusable at lunchtime?”
Wi-Fi radios operate on license-free frequencies. This
allows anyone to set up a Wi-Fi network, but it also means
Wi-Fi needs to compete with other devices using these
frequencies. Such devices include microwave ovens,
wireless video cameras, Bluetooth®, baby monitors, home
automation systems, radar and more.
Microwave ovens, for example, are wide-band inter-
ferers from the Wi-Fi point of view. A microwave oven
leaks radiation on all 2.4 GHz Wi-Fi channels. One oven
does not kill the frequency entirely, but it has a significant
impact on capacity for the nearby devices. Many wireless
video cameras, on the other hand, utilize only a few of the
2.4 GHz channels, leaving room for a smart Wi-Fi engineer
(or a smart Wi-Fi infrastructure) to cope with the situation
by adjusting the channel separation of the Wi-Fi network.
Interfering devices can easily be detected by a spec-
trum analyzer (Figure 4). There are two types of spectrum
analyzers, and a wireless engineer would ideally have
both tools available:
u A portable spectrum analyzer that can be connected
to a laptop via USB. It is easily portable for performing
site surveys and troubleshooting in desired locations.
The downside is the spectrum analysis is periodic in
nature at best, not constant.
u A spectrum analyzer built into APs allows for constant
monitoring of the spectrum once the network is up
and running. However, the measurements are measured
at the ceiling level, not where the users are.

STANDARD FREQUENCY TYPICAL AP SIGNIFICANCE

MAX DATA RATE
802.11 (legacy) 2.4 GHz 2 Mb/s The first WLAN standard

802.11b 2.4 GHz 11 Mb/s Made WLAN popular

802.11a 5 GHz 54 Mb/s The first 5 GHz standard

802.11g 2.4 GHz 54 Mb/s Faster speeds on 2.4 GHz

802.11n 2.4/5 GHz 450 Mb/s Wider channels, more spatial streams = more speed.

802.11ac 5 GHz 1.3 Gb/s Even wider channels, even better modulation = even more speed.

TABLE 1: Properties of selected IEEE 802.11 standards.

10 u ICT TODAY
 Are You at Risk?
TIA / ETL
Component Rated? + Solderless PCB? + 50 Gold Rated? + PoE Rated?
®

UL® / cUL® Listed No Intermittent Failure 50 Micro-inch Gold IEEE 802.3at PoE Plus

Rear 110 -Type IDC 750 Plug-Jack Mating Cycles 100+ Re-termination RoHS Compliant

Not complying with FCC or TIA standards might prevent you from getting paid. Ask your data connectivity
vendor these questions:

Are the patch panels TIA /ETL “Component-Rated?” Are the patch panels made with solderless PCB?
A component rating is a tougher standard than the “permanent link” rating Punch-down termination requires up to 50 pounds of harsh force. This force
some manufacturers claim to have. can break the solder joint between 110 and PCB. We only use solderless 110s.

Are the patch panels FCC part 68 compliant? Are the patch panels PoE rated?
FCC specifies 50 micro inch gold on contact. Every time you plug in, you shave If you are installing a PoE system, a PoE rating is required. Without it, you
off a little gold plating. 50 Gold® will cover you for up to 750 insertion cycles. risk damaging your equipment.

With all these enhancements, ICC patch panels still cost 50% less than the big brands. If you are still paying for
big brands, we need to talk!

24-Port, 1 RMS 48-Port, 2 RMS 12-Port Wall Mount

Available at these Premise Cable Stocking Distributors:

888-ASK-4ICC
icc.com csr @ icc.com
© Copyright 2016, ICC. May/June 2016 t 11

mpr-0445 - ICT JOURNAL MAY-JUNE 2016.indd 1 4/1/2016 8:55:13 AM

On and Off the Network Quickly
When it comes to Wi-Fi, one device talks at a time
per channel. That means when the AP is sending data
to one of the client devices, everyone else is waiting.
Then, a single client device might send data to the AP,
while everybody else waits for their turn. This is why it
is ideal to get the devices on and off the airwaves very
quickly. The better this “data rate” of transmission, the
faster the job is done.
High data rates require high signal-to-noise ratio
(SNR), i.e., high signal strengths and low noise combined.
In addition, the network simply needs to have a fairly
modern wireless network infrastructure that supports
high data rates to begin with. The same is true for client
devices. The newer IEEE 802.11n and 802.11ac standards
have introduced multiple input multiple output (MIMO,
which sends multiple data streams simultaneously between
the client and the AP) and enhanced modulation schemes
that allow multiple times the data rates compared to older
802.11a/b/g standards.
At times, the client devices may be connected to APs
that are farther out, for whatever reason, and operating at
low data rates. Similarly, old client devices (802.11a/b/g)
only support low data rates. One possible solution is to
disallow low data rates from the APs. This means high
data rates, at the cost of losing support for very old client
devices, and client devices will not be able to connect far
away from the APs.
It is important to note that data rate is not the same
as throughput. Rather, it is the theoretical maximum at
which the client and APs are able to communicate.
Throughput is the amount of actual data being transmit-
ted. For example, high data rates between a single client
and an AP may be achievable at times even on a badly
designed network, but the total simultaneous throughput
of all the client devices will suffer dramatically if proper
coverage and channel planning has not been done.
High Capacity Considerations
How should the demand for high wireless capacity
be addressed? Following network design best practices is
a good start. Here is an example of a process for designing
high-capacity Wi-Fi networks:
1. Determine the business needs: What are the problems
to be solved by utilizing Wi-Fi?
2. Specify technical requirements: How many users are
there? What kind of devices, and how many, do they
have? What kind of applications are they using, and
how often?
12 u ICT TODAY
3. Calculate the need for infrastructure: Based on the
number or client devices, their bandwidth require-
ments and capabilities, calculate the number of
required APs to support the usage.
4. Predictive design: Map out the exact locations,
configurations and antennas for the APs. Get
simulated heat maps to ensure sufficient network
coverage and SNR, and minimize channel overlap
based on the walls of the building.
5. Network deployment: Install and configure the
wireless infrastructure and the wired infrastructure
to support it.
6. Verification site survey: Walk around the facility
with a site survey tool to map out the network
and ensure sufficient Wi-Fi coverage and perform-
ance. A passive survey is always required; preferably,
this is complemented by an active survey, a
throughput survey, and a spectrum analysis survey.
7. Maintenance: Every network needs maintenance. If it
is well designed, it needs very little maintenance.
A badly designed network requires constant trouble-
shooting and fine-tuning.
Conclusion
To design a high-quality Wi-Fi network that supports
the increasing demand for capacity:
u Pay attention to designing the Wi-Fi network to avoid
constant troubleshooting and network modifications.
A predictive network design (simulation) accounting
for wall materials is a good start.
u Calculate the need for capacity based on the number
of users, their devices and their applications.
u Achieve high data rates by ensuring high signal
strength, utilizing fairly modern network gear,
minimizing noise and verifying the network with
a site survey.
u Maximize airtime by minimizing channel interference
and the number of service set identifiers, or SSIDs, as
well as by preferring the 5 GHz frequency. t
AUTHOR BIOGRAPHY: Jussi Kiviniemi oversees the Wi-Fi Design
Tools business at Ekahau. An industry pioneer and a familiar face in the
Wi-Fi (WLAN) space, he has been featured in publications from IEEE papers
to TheStreet.com, and presented at dozens of conferences, including BICSI’s
largest events, SCTE, Wireless LAN Professional Conferences, and CWNP
(Certified Wireless Networking Professional) conferences. Jussi holds an
M.S. in industrial engineering and management. He can be reached at
jussi.kiviniemi@ekahau.com.
 Is Your Network Running
DL_BICSI mag_fp ad.pdf 1 4/4/16 2:50 PM

onIs Old
YourTechnology?
Network Running
on Old Technology?
Is Your Network Running
on Old Technology?

Upgrade Easily with Air-Jetted Fiber Today

Y

CM

Upgrade Easily with Air-Jetted Fiber Today

Fiber optics is the next generation for video and IP routing.
MY

CY

Greater bandwidths are required to meet today’s expanding needs.

Fiber optics is the next generation for video and IP routing.
CMY

Greater bandwidths are required to meet today’s expanding needs.

K

• FuturePath (bundled MicroDucts) •

eABF networks solve your bandwidth needs by providing a scalable, flexible and
Upgrade Easily with Air-Jetted Fiber Today
• Up to 24 pathways/bundle
• FuturePath (bundled MicroDucts) smart solution. eABF networks offer a reliable, easy-to-install fiber optic network
• eABF Fiber eABF
• Up to 24 pathways/bundle
• Up to 96 fibers/pathway
with anetworks
physical solve yourand
pathway bandwidth needs
high density by providing
fiber engineereda for
scalable, flexible and
air-jetting.
smart solution. eABF networks offer a reliable, easy-to-install fiber optic network
• eABF Fiber
• Up to 96 fibers/pathway
Fiber optics
with is thepathway
a physical next generation for video
and high density fiberand IP routing.
engineered for air-jetting.
™
GreatereABF is a partnership
bandwidths between
are required AFL and
to meet Dura-Line.
today’s expanding needs.
eABF solutions
™ are designed to meet your network’s current bandwidth needs,
• FuturePath (bundled MicroDucts)
• Up to 24 pathways/bundle
Scalable
eABF networks
eABF is a partnership between AFL and Dura-Line.
solvescalable
and are easily to accommodate
your bandwidth needs morebyfiber in the future.
providing No needflexible
a scalable, to and
splice, connectorize,
eABF solutions and
are test fiber
designed tothat
meetyou're
your not ready
network’s to use
currenttoday!
bandwidth needs,
• eABF Fiber Scalable
smart solution. eABF networks offer a reliable, easy-to-install
and are easily scalable to accommodate more fiber in the future. No need to
with a physical pathway and andtesthigh density fibernotengineered
fiber
for
optic network
air-jetting.
• Up to 96 fibers/pathway splice,
A well connectorize,
designed, flexible network fiber that you're
architecture with eABF ready to use easier
enables today!
Flexible customization than both conventional conduit and traditional cabling solutions.
Ask about
A well the benefitsflexible
designed, of a pathway!
network architecture with eABF enables easier
Flexible eABF customization
™
is a partnership between
than both conventional AFL and
conduit andtraditional
Dura-Line.cabling solutions.

Smart needed
Scalable
Save Ask about
money onthe benefits of
installation, a pathway!
fiber and investment in technology
eABF solutions are designed to meet your network’s current bandwidth
and arenow. eABF
easily solutions
scalable to can be quickly more
accommodate and economically
that isn’t needs,
expanded
fiber in the future. at anyto
No need
Smartforsplice,
time with
Save
today’s
minimal
money
neededneeds
onlabor and disruption
installation,
and defer
now. eABF
fiber
costs can
solutions
andto your facility.
investment
untilbegrowth
quicklydemands!
in It’s just
connectorize, and test fiber that you're not ready to use today! isn’tsmart
technology to
that
and economically expanded at any
pay

time with minimal labor and disruption to your facility. It’s just smart to pay
A well designed, flexible network architecture with eABF enables easier
Contact usFlexible
today to learn how to air-jet fiber! for today’s needs and defer costs until growth demands!
customization than both conventional conduit and traditional cabling solutions.
Ask about the benefits of a pathway!
Contact us today to learn how to air-jet fiber!
Smart Save money on installation, fiber and investment in technology that isn’t
needed now. eABF solutions can be quickly and economically expanded at any
time with minimal labor and disruption to your facility. It’s just smart to pay
for today’s needs and defer costs until growth demands!800.847.7661
www.duraline.com/enterprise
800.847.7661
Contact us today to learn how to air-jet fiber! May/June 2016
www.duraline.com/enterprise
t 13
Data Center Infrastructure
Management Solutions:
HOW TO HIT
A MOVING
TARGET

By Bob Potter For some 20 years, data center design evolved more or less
linearly—that is to say, as expected and without significant
deviation. Data centers grew larger, added racks and
increased density, and the power and cooling systems they
required evolved accordingly. This is the story of the traditional
enterprise data center.

14 u ICT TODAY
 As the industry matured, energy
use and costs skyrocketed, end-user
demands intensified, and new tech-
nologies emerged. Organizations
started to question some long-
standing assumptions about what
a data center is and could be:
How much computing will be
needed five years from now?
Is there justification for build-
ing a facility to house 500
racks when only 100 are
needed today?
Is it affordable to build for
500 when, someday, 1,000
might be required?
Whatever is decided, how can
energy costs be managed?
Those questions, along with
the increasing complexity of the
traditional data center model, drove
the development of data center
infrastructure management (DCIM).
Early DCIM solutions were designed
with enterprise deployments in mind,
focused on reining in the mostly
unchecked waste associated with
the overprovisioned facilities of the
early 2000s. DCIM would provide
a tool to track and manage assets
across those increasingly virtualized
environments.
But an interesting thing
happened on the way to DCIM
delivery—the data center evolved
again. Server capacity utilization
increased, while energy efficiency
and visibility across systems degener-
ated. The industry started to explore
and eventually embrace new arch-
itectures and approaches. Some
organizations abandoned owned
facilities altogether and moved to
cost-efficient colocations or the cloud.
DCIM Evolution and the
Modern Data Center
The earliest DCIM products
launched in a new world in which the
benefits of unprecedented visibility
and control across the physical
layers of the network were clear and
unquestioned, but the early adopters
largely were large traditional data
centers. DCIM immediately became
and remains a valuable solution in
cloud, colocation and hyperscale
environments. However, traditional
data centers were far from extinct.
It would be too strong a statement
to say the first DCIM solutions were
outdated on arrival—after all, those
early adopters enjoyed better asset
management, increased efficiency
and improved availability. But almost
as soon as it arrived, DCIM had to
change to better meet the needs of
the evolving information ecosystem.
As DCIM platforms have become
more sophisticated and the benefits
more pronounced, more data center
and IT managers are investigating
DCIM for their networks—even if
those networks do not fit the big
enterprise data center model.
What they have found, especially
over the last two years, is a changing
DCIM landscape, with scalable
options that allow organizations to
build a DCIM solution that fits their
data center and can grow as the
organization grows.
Today’s data center is no single
thing. There are many traditional
enterprise facilities, but they are
more complex than ever before.
Although conventional AC power
architectures remain the primary
choice in these environments,
alternative architectures are
emerging—everything from high-
voltage DC to telco-tested modular
DC power to hybrid models.
Acceptable temperatures are rising,
but intelligent thermal management
is critical to maintaining availability
and optimizing energy use. Demands
for speed and security are driving
computing closer to the edge. This
is forcing providers to rethink
network architectures and vendors
to reduce their equipment footprint
and increase visibility and remote
management capabilities.
Applying DCIM to
Distributed Networks
Consider distributed data
centers—an increasingly common
model. They often have multiple
small computing modules spread
across different locations, providing
local computing and storage while
still networking with each other
and with the small data center at
the network’s center. Individually,
these are simple IT resources—often
just a single rack of equipment.
But collectively, these are complex
networks with significant manage-
ment challenges that the earliest
DCIM solutions were never designed
to address. So, just as the data
center has evolved, so has DCIM.
Today’s best solutions are flexible,
scalable and modular, suitable for
immense data centers, cloud facilities,
colocation pods or those distributed
data centers with hundreds of
scattered computing nodes. In fact,
because of the disparate nature of
today’s data centers, effective DCIM
is more important and more valuable
than ever before.
May/June 2016 t 15
As data center infrastructure management (DCIM) platforms have become
more sophisticated and the benefits more pronounced, more data center
and IT managers are investigating DCIM for their networks—even if
those networks do not fit the big enterprise data center model.
Take the case of Cambridge
University in England, a school
that measures its history not in
decades but in centuries. As the
campus and various departments
have grown, evolved and embraced
the IT revolution over the past 30
years, separate micro networks
have popped up across the campus.
Integration with the school’s main
data center was scattered at best. As
a result, the campus ended up with
more than 200 server rooms serving
120 departments, and most of them
operated independently.
This is not especially unusual,
but it is exceptionally inefficient.
These separate, independent IT nodes
included equipment from multiple
vendors, eliminating the benefits
of consolidated management,
standardized service delivery and
improved security and availability.
The result: some 200 loosely
connected server rooms with little or
no consistency across equipment or
operational practices.
Eventually, the university
prioritized improved visibility and
management of these scattered
computing clusters with some
clear objectives: to optimize the
performance and efficiency of its IT
systems and reduce operational costs.
The school consolidated several of
the IT rooms in a single modern
data center and unified the new
16 u ICT TODAY
facility and remaining distributed
computing sites under a single
DCIM system. Although originally
conceived with a traditional data
center model in mind, applying
DCIM to improve visibility and
control of a web of hard-to-see IT
assets is exactly what the technology
was developed to do. The various
locations and disconnected nature
of the Cambridge facilities just
added another layer of complexity.
Ultimately, the desire was the same
as it is for any DCIM customer: to see
and control multiple assets—not just
servers, but every component across
the network—from a single location.
Cambridge deployed a DCIM
solution that enabled efficient
management of multi-vendor IT,
power and cooling resources. The
system configures and organizes data
from all of Cambridge’s computing
facilities and translates it into a
unified, actionable language, helping
the university achieve higher
efficiency—with a projected power
usage effectiveness (PUE) of 1.2 and
improved performance. Simply put,
it has made everything the university
does around its IT systems smarter.
The Stages of DCIM
There are four stages to
effective DCIM deployment, each
addressing critical questions which,
when answered, improve network
performance and efficiency. One of
the great benefits of today’s modular
DCIM platforms is the ability to
deploy at any of these four stages,
making each of them a potential
entry point for tailored DCIM
implementation instead of hurdles
to be cleared on a long road to data
center optimization.
STAGE 1:
Data Capture and Planning
What and where are assets in
the data center?
How are they interconnected?
Is there space, cooling and power
to meet future needs?
How can assets be efficiently
commissioned and
decommissioned?
Benefits: Improves planning and
provides data needed to improve
efficiency.
STAGE 2:
Monitor and Access
How are assets operating?
Are there real-time notification
of alarms and alerts?
How does a server get back up
and running?
Can planning tools be populated
with actual performance data?
Benefits: Provides early warnings that
minimize service requirements and
ensure availability.
STAGE 3:
Analyze and Diagnose
How can the life of the data
center be extended?
How can mean time to repair
(MTTR) be reduced?
How can infrastructure be
synched with virtualization
automation?
Benefits: Reduces MTTR and
service requirements, manages
infrastructure capacity, and tracks
performance.
STAGE 4:
Recommend and Automate
How can potential failures be
anticipated so that compute
and physical loads can be
automatically shifted to
eliminate downtime?
How can efficiency be opti-
mized across my data center?
Benefits: Ensures availability and
optimal performance.
Finding the Right Fit
The key to answering all
these questions is finding a DCIM
solution that helps the data
center become a business asset
instead of a business expense.
DCIM technologies are not one-
size-fits-all solutions to whatever
is ailing the data center. Right-
sizing is important, and finding a
vendor capable of matching the
right solution to your data center
is critical to optimizing your IT
investment. Choosing a vendor
that has the right fit today but no
option for growth is as inefficient
as choosing one who only offers
large, enterprise solutions you
hope to grow into—someday. The
single biggest development in
DCIM in the last two years—and
the most significant differentiator
between today’s solutions—is
scalability.
That capability was one of
the critical data points in recent
DCIM evaluations from Gartner
and IDC. Gartner’s 2015 DCIM
Magic Quadrant and the IDC
MarketScape: Worldwide Data
Center Infrastructure Management
2015 Vendor Analysis evaluated
current DCIM offerings on
many different elements of
capabilities and segment strategy,
including scalability and modular
capabilities. Those capabilities
allow businesses to choose DCIM
bundles that meet immediate
needs while allowing for future
growth, thus ensuring IT remains
a critical business asset. t
AUTHOR BIOGRAPHY: Bob Potter
is a Senior Product Marketing Manager
for Emerson Network Power’s Software
business unit. Bob joined Emerson in
2011 as Product Marketing Manager for
the Trellis Platform software solution
and has managed the marketing efforts
from initial launch in 2012 to date. Prior
to joining Emerson, Bob held software
product management and marketing
roles with Stanley Black & Decker and the
McGraw-Hill Companies. He led several
successful software product launches for
these organizations. Bob holds a B.S. and
an M.B.A. in Marketing from Grand Valley
State University. He can be reached at
bob.potter@emerson.com.
May/June 2016 t 17
SPEAKER PLACEMENT
AND WIRING GUIDELINES
FOR PUBLIC ADDRESS/PAGING SYSTEM/AV
Public address (PA)/paging/AV systems have been a
part of the information and communications technology
(ICT) industry for more than 90 years. Based largely on
technical developments and research conducted within
the telephone and radio industries by organizations
such as Bell Labs and RCA early in the 20th century,
the invention of the vacuum tube and the loudspeaker
allowed for the commercial deployment of these systems
18 u ICT TODAY
following World War I. Early applications included movie
theaters, concurrent with the development of motion
pictures with soundtracks, and outdoor sports stadiums.
Further developments in later years expanded
the reach of these systems into an increasing number
and variety of applications, both public and private.
These include locations such as schools, offices, retail
establishments and transportation facilities.
By Robert B. Hertling Jr., RCDD, OSP
While technological developments
have continuously changed the config-
uration and components of these
systems over the years, the basic concept
and purpose of these systems has not
changed. Most PA/paging/AV systems
can be considered to be “distributed
reinforcement” sound systems in that
they provide real-time amplification of
an audio source (e.g., live or recorded
audio announcements or media audio) to
listeners in a served area using a number
of individual speakers. This basic, but key,
concept is the foundation of the successful
design and installation of these systems.
In his book Sound System Engineering,
Don Davis writes, “Successful reinforce-
ment systems must be loud enough
(sufficient acoustic gain), possess clarity
(provide a low percentage of articulation
loss of consonants in speech), and cover
the listeners with uniformity, while
avoiding the coverage of areas devoid
of listeners.”
Regardless of the size or specific use
of these systems, speaker spacing and
location are a key part of ensuring that
a PA/paging/AV system provides the
required degree of coverage. Poor spacing
or bad location selection can result in a
system that provides marginal or no useful
information to listeners.
This article presents a discussion of
basic concepts and key concerns related
to system design and installation, speaker
placement and wiring guidelines. It is not
intended to be inclusive of all design and
installation considerations, since each
system location will have differing require-
ments, unique constraints, and specific
client/user expectations. Additional
specialized professional acoustical and
communications engineering support is
often required for these situations.
SPEAKER CHARACTERISTICS
Speakers are considered to be transducers, in that they
convert electrical energy into mechanical energy that results
in the generation of sound. Most PA/paging/AV systems use
two basic speaker types: coaxial cone and horn.
While there are several methods of performing this energy
conversion, most typical coaxial cone and horn speakers
utilize electromagnetic drivers to transform the electrical
energy input to a mechanical energy output. They can be
considered to be a specialized form of linear electric motor.
COAXIAL CONE SPEAKER TYPICAL COMPONENTS:
FRAME—provides mechanical rigidity and
mounting for all components.
VOICE COIL—performs the conversion from
electrical energy to magnetic energy.
POLE PIECE—of magnetic material,
is attracted and repelled by magnetic
fields generated in the voice coil.
DIAPHRAGM—usually of paper or plastic, the mechanical
movement of this item by the pole piece produces sound.
SPIDER—connects the diaphragm to the frame and to the
pole piece for stability while still allowing the movement
required to produce a sound signal output.
HORN SPEAKER TYPICAL COMPONENTS:
FRAME—provides mechanical rigidity
and mounting for all components.
VOICE COIL—performs the
conversion from electrical energy
to magnetic energy.
POLE PIECE—made of magnetic material, it is attracted and
repelled by magnetic fields generated in the voice coil.
DIAPHRAGM—usually made from paper or plastic, the
mechanical movement of this item by the pole piece
produces sound.
SPIDER—connects the diaphragm to the frame and the
horn and to the pole piece for stability while still allowing
the movement required to produce sound.
HORN—usually a double- or, in some cases, a triple-folded
arrangement, aims and focuses the sound in an arranged
directed pattern.
May/June 2016 t 19

FUTURE OF THE INDUSTRY
Visit bicsi.org/volunteer to apply to sit on these committees and help shape the future of ICT.
WRITING OPPORTUNITIES
MANUALS:The Technical
Information and Methods (TI&M)
Committee is responsible for
writing and updating the BICSI
design and installation technical
manuals. The committee is
composed of SME members
and other non-committee
SMEs, from all over the world,
who assist in writing, performing
reviews, making suggestions
for applicable revisions and
updates, and agreeing
on changes through a
consensus-based process.
HELP GUIDE THE
Experts (SMEs) Needed
Manuals and Standards.
STANDARDS:The Standards
Committee participates
in the development of
telecommunications-related
standards, as well as notifies
BICSI’s membership of
changes within applicable
standards that may affect
them. Several different
subcommittees make up the
Standards Committee.
Subject Matter
to Contribute to BICSI
OTHER:Write articles
for BICSI’s premier trade
publication, ICT Today. Writers
share vendor-neutral, relevant
and authoritative insight on
ICT design and installation
practices and solutions to
help our readers expand their
scope of work and enhance
their professional stature. It’s
an excellent opportunity to
share your knowledge and
expertise with your industry
peers. Send your article
ideas to icttoday@bicsi.org!
 Since speakers have an electrical input which
produces a mechanical output (namely, sound), there are
both electrical and mechanical (acoustical) characteristics
that are functions of speaker performance and which
need to be taken into consideration in design and
installation activities.
The acoustical parameters that are important
considerations in speaker selection and spacing/
location include:
u Audio Frequency Response: Measured in
Hertz (Hz), frequency response is the range of audio
frequencies that the speaker can faithfully reproduce.
For successful reproduction of speech in a PA/paging
system, the accepted minimum frequency response
of all components within the system is typically
350 Hz–5 kilohertz (kHz). For AV systems, the
accepted minimum frequency response of all
components within the system can be as broad as
20 Hz–20 kHz.
u Sensitivity: Measured in decibels (dB) and sound
pressure level (SPL). This is the on-axis (i.e., directly
in front of or below the speaker) loudness produced
by the speaker in dB SPL measured at a specific
distance—usually ≈1 meter (m [3.3 feet (ft)]) with a
specified electrical power input (usually 1 watt [W]).
SPL can range from 0 dB (threshold of hearing
for a typical person) to 120 dB (threshold of pain
for a typical person). As an example, a subway train
entering a station typically can generate 90 dB SPL
measured at ≈6.1 m (20 ft). An average person’s
voice at a conversational level generates 70 dB SPL
measured at ≈0.3 m (1 ft).
SPL represents sound energy intensity—what is
commonly referred to as loudness. An increase of
10 dB SPL is perceived by a typical listener as
doubling the volume of the sound. It is important
to note that SPL is an acoustical/mechanical
measurement, not an electrical measurement.
u Dispersion angle: Measured in degrees, this is
the angular value within which the SPL is not more
than 6 dB below the on-axis level (the sensitivity
level) for the speaker’s overall frequency response or
a specific frequency specified by the speaker
manufacturer.
For the purposes of this discussion, the following
parameters for a typical ceiling-mounted 8-inch (in)
coaxial speaker will be used:
u Frequency response—60 Hz–16 kHz
u Sensitivity—97 dB/1 W/1 m SPL
u Dispersion angle—50 degrees off axis
ACOUSTIC CONSIDERATIONS
Acoustic considerations for speaker spacing/
location include:
u Space dimensions and configuration (length,
width, height, circular, rectangular, etc.) for the
area to be covered.
u Ambient noise (ranges from low, as in an office
environment, to high, as in an industrial
environment) within the area to be covered.
u Surface characteristics (reflective—concrete, ceramic
tile and similar surfaces or absorptive—carpet, fiber
ceiling tile and similar surfaces) within the area to
be covered.
For the purposes of this discussion, the following
parameters will be used:
u Space dimensions (includes length, width and
ceiling height) and configuration (rectangular)
u Ambient noise (medium to high)
u Surface characteristics (reflective)
DETERMINING COVERAGE AREAS
To determine coverage areas, the designer should:
u Obtain space measurements and prepare plan and
elevation views.
u Note existing surfaces and finishes and identify areas
not to be covered.
u Identify special conditions, such as open archways
to stairwells or other spaces, or abrupt changes in
dimensions, such as ceiling heights.
u Calculate the base coverage area for an individual
speaker based on the manufacturer’s specifications.
Figure 1 on the next page provides the methodology
for calculating the base coverage area, using basic
trigonometry and the dispersion angle and sensitivity
values noted earlier. While the calculations shown here
are being done by hand in order to demonstrate the
individual steps involved, there are commercial software
May/June 2016 t 21
 applications available to perform these calculations
1 watt input
and the subsequent steps involved in determining
speaker placement.
50 degree Once the speaker’s base coverage area has
dispersion
angle SPEAKER
been determined, the next step is to determine
the location-specific coverage area by including
1 meter
Radius (r) = tan50x1 two additional measurements—distance from
= 1.1918 meters
speaker to listener ear height and desired SPL at
97 dB SPL listener ear height—into the calculations. Assume
a 12-ft ceiling height from the floor, a 5-ft listener
ELEVATION VIEW PLAN VIEW
ear height, and a minimum of 75 dB SPL and
maximum of 95 dB SPL at listener ear height.
FIGURE 1: Methodology for calculating the base coverage area. Note that, in many cases, the level of 95
dB SPL at listener ear height is the maximum
1 watt input allowable in order to avoid the potential for
hearing damage to people present within the
covered area.
50 degree
dispersion There are two important concepts to remember
angle SPEAKER
about SPL values:
2 meters u If the reference distance for the initial SPL
Radius (r) = tan50x2
(6.6 ft)
= 2.3836 meters
calculation is doubled, the SPL will decrease
(7.8 ft)
91 dB SPL by 6 dB (e.g., base coverage area SPL at ≈1 m
[3.28 ft] is 97 dB at 1 W, at ≈2 m [6.6 ft] the
ELEVATION VIEW PLAN VIEW SPL will be 91 dB at 1 W). This principle is
also known as the Inverse Square Law.
FIGURE 2: Calculating the location-specific coverage area for an individual speaker u If the reference input electrical power for the
using data for space dimensions, ambient noise and surface characteristics initial SPL calculation is doubled, the SPL will
increase by 3 dB (e.g., base coverage area SPL
at ≈1 m [3.28 ft] is 97 dB at 1 W, at ≈1 m
[3.28 ft] the SPL will be 100 dB at 2 W).

For this example, the measured ambient noise

within the area to be covered ranges from 55 to
60 dB SPL. The SPL at listener ear height should be
15-20 dB above the ambient noise level.
Calculate the location-specific coverage area
for an individual speaker using data for space
dimensions, ambient noise and surface character-
istics. Figure 2 illustrates this calculation process.
Once this step is complete, review the calculations.
u Do the results provide the necessary SPL level at
listener ear height? If they do, the next step is to
SQUARE PATTERN HEXAGONAL PATTERN determine the speaker layout within the space.
u If they do not, there are additional steps
which must be taken before proceeding with
FIGURE 3: Basic patterns for speaker layout. the layout:

22 u ICT TODAY
 BICSI’s NEW Corporate
Connection Program

membership
membership
MEMBERSHIP
Offering multiple tiers of membership—select the option that will best fit

training
your organization.

training
TRAINING

discounts
Opportunities for training your employees and promoting your company.

discounts
DISCOUNTS
Each tier includes discounts and benefits tailored to let you get the most
out of your membership.

IS YOUR COMPANY CONNECTED?

Learn more and sign up today
at bicsi.org/connected.

May/June 2016 t 23
 u Increase or decrease the height of the speaker
above the floor.
u Increase or decrease the input electrical power
to the speaker.
u Select a different speaker—usually a speaker
with a smaller dispersion angle can provide a
SPEAKER SPEAKER greater SPL output at ≈1 m (3.28 ft) than one
with a wide dispersion angle. If this is the
case, the calculations must be redone before
proceeding with the speaker layout.

SPEAKER LAYOUT
FIGURE 4: Edge-to-edge spacing method. Once the final coverage area for an individual
speaker has been determined, the next step is to
evaluate and select a speaker layout methodology.
Two basic patterns exist as shown in Figure 3 on
page 21: square and hexagonal.
The choice of pattern depends on the best fit
between the space dimensions and the speaker
coverage areas. Also, the pattern orientation can
be rotated as needed to fit the shape of the space.
SPEAKER SPEAKER

Three basic spacing methodologies exist:

1. Edge-to-edge: The spacing distance is equal
to 2r (r = radius of the coverage area). Edge-
to-edge requires the least number of speakers
to cover a space. However, it will leave
FIGURE 5: Minimum overlap method. some gaps in coverage. It is not recommended
for locations with poor acoustics or significant
background noise. This methodology is shown
in Figure 4.

2. Minimum overlap: In this case, the spacing

distance is equal to r times 1.414 for speakers
utilizing a square pattern and r times 1.732
SPEAKER SPEAKER for speakers utilizing a hexagonal pattern.
Minimum overlap spacing requires more
speakers than the edge-to-edge spacing
methodology. While it will cover some gaps
in the coverage area as compared to the
edge-to-edge spacing methodology, it may
FIGURE 6: Edge-to-center method. not be sufficient in many cases and may still
leave areas where the audio may not be heard
or is unintelligible in areas with poor
acoustics or significant background noise.
This methodology is shown in Figure 5.

24 u ICT TODAY
3. Edge-to-center: This method utilizes the highest
speaker density commonly used for PA/paging
systems. It is the best methodology for areas with
poor acoustics or significant background noise.
The spacing distance is equal to r. It requires an
additional quantity of speakers over that required
for the minimum overlap and edge-to-edge spacing
methodologies. This methodology is shown in
Figure 6.
ELECTRICAL CONSIDERATIONS
Once the speaker spacing and location issues have
been addressed in the design, the next step is determin-
ing the electrical requirements and constraints. For
systems using a voice over Internet protocol (VoIP)/
Ethernet-based distribution methodology, these factors
can include horizontal cabling length limits (100 m
[295 ft]) per the governing TIA standards and powering
availability for the individual speaker locations. Powering
considerations can include Power over Ethernet (PoE)
limits and/or availability of local 120VAC power.
For systems using a constant voltage distribution
methodology, factors to be considered can include
the following:
u Audio power requirement determination:
These systems utilize nominal voltage level audio
output circuits from audio power amplifiers.
Typically this voltage is 25 or 70.7 volts (V), but
in some instances it could be 100 V or higher. The
speakers are wired in parallel to the audio output
circuits via multi-tap matching transformers at
each speaker location. The matching transformers
are used to match the impedance of the speaker
voice coil (typically 8 ohms) to the high impedance
of the constant voltage audio output circuits of
the amplifiers and allow, through the multi-taps
on the primary of the transformer, selection of
the power in watts to be provided to the speaker.
Matching transformers can be purchased with taps
as low as ¼ watt (W) up to values as high as 15 W
along with various levels in between these values.
The selected transformer must always match the
speaker voice coil impedance and must not allow
the power to exceed the speaker manufacturer’s
maximum; otherwise, damage to the speakers and
other system components, such as amplifiers, may
result. Calculating the audio power required involves
obtaining all of the individual speaker power require-
ments based on the matching transformer tap
settings and adding them together (e.g., 25 speakers
each tapped at 1 W = 25 W).
u Audio circuit configuration: Determine the
number of circuits required to connect the speakers
to the amplifiers. Zoning requirements, separation of
spaces within the same zone, circuit redundancy and
pathway/raceway configuration are just some of the
factors to be considered in determining the number
of circuits required.
Next, assign speakers to each circuit. This is best
done by utilizing one-line or riser diagrams with
each speaker uniquely identified to its location on
the plan and elevation drawings. Finally, ensure that
circuit connections are polarized (+/-) correctly. Even
though audio circuit connections and components
such as matching transformers and speakers are
part of an alternating current (AC) circuit in a
constant voltage distribution methodology, the
polarity directly determines the phase relationship
of both the electrical and acoustical signals within
the system. Incorrect polarity can create an “out of
phase” condition where speaker outputs can interfere
with or, in extreme cases, cancel each other, resulting
in reduced or no sound levels and/or distortion.
u Audio circuit sizing: Usually, a minimum of
16 AWG wire is specified for audio output circuits
between the amplifier outputs and the speakers.
In a 70.7 V PA system, 16 AWG wire is limited to a
maximum safe current of 6 amperes (A), resulting
in a maximum power capacity of 420 W at a
maximum distance of 90 ft, assuming a 0.5 dB (12.5
percent) line loss. In some cases, wire size may have
to be increased to meet the power and/or distance
limitations within a particular circuit (Table 1).
u Amplifier loading: According to manufacturer
recommendations/best practices, the connected
load should not exceed 80 percent of the amplifier
power rating—for a 100 W rated amplifier, the total
connected load should not exceed 80 W, including
circuit losses.
May/June 2016 t 25
 WIRE OHMS MAX MAX MAX MAX MAX MAX MAX MAX MAX
SIZE PER SAFE SAFE LENGTH LENGTH LENGTH LENGTH LENGTH LENGTH LENGTH
1000' AMPS POWER (FT) AT (FT) AT (FT) AT (FT) AT (FT) AT (FT) AT (FT) AT
LOOP (R) (I) (W) 10W 15W 20W 30W 40W 60W 100W

#15 8.0 6 420 3600 2400 1800 1200 900 600 370

#14 5.2 15 1000 5300 3800 2800 1900 1400 950 560

#12 3.2 20 1400 9100 6200 4600 3100 2300 1500 910

Courtesy of Altec Lansing

#10 2.0 25 1750 9900 7300 5000 3700 2500 1450

#8 1.28 35 2450 7800 5700 3900 2280

TABLE 1: Length of two-wire 70.7 V line delivering various values of power at 0.5 dB (12.5 percent) loss.

CODE AND AUTHORITY HAVING
JURISDICTION (AHJ) CONSIDERATIONS
In all cases, system wiring must comply with all
applicable codes and standards. In the United States,
NFPA 70 and the National Electrical Code® (NEC®)
Article 640 (Audio Signal Processing, Amplification,
and Reproduction Equipment) contains the primary
governing requirements for PA/paging/AV systems to
be enforced by the Authority Having Jurisdiction (AHJ).
Article 640 also contains references to Article 725 (Class
1, Class 2 and Class 3 Remote-Control, Signaling and
Power-Limited Circuits).
At this point, a question often arises: Are these NEC
articles relevant to system implementation, regardless
of whether it utilizes a VoIP/Ethernet architecture or a
constant voltage distribution methodology? The answer
is: it depends on the AHJ’s interpretation. The NEC
currently does not differentiate between the two types
of systems. Article 640 does allow the use of Class 2 or
Class 3 power-limited wiring as defined in Article 725,
provided the amplifier assemblies are listed and marked
for use with Class 2 or Class 3 power-limited wiring—
this is typical for amplifier assemblies having output
power no greater than 100 W, in order to meet the
supplied power limits defined in Article 725.
Article 725 specifically prohibits audio circuits using
Class 2 or Class 3 power-limited wiring to occupy the
same cable or raceway as other Class 2 or Class 3 power-
limited circuits.
Many AHJs also prohibit audio circuits using Class
2 or Class 3 power-limited wiring from occupying the
same cable or raceway with communications circuits as
defined in Article 800. As a result, these systems may
be required to utilize cabling and pathways that are
partially or totally independent of other ICT cabling
and infrastructure within a premises.
CONCLUSION
This article has attempted to identify basic concepts
and considerations involved in speaker placement and
wiring guidelines for PA/paging/AV systems. As stated,
it may be necessary to obtain specialized professional
acoustic and communications engineering support
for large systems and/or those with unique coverage
requirements. However, the end result should be the
same: a system which provides the required degree
of coverage and presents useful information to the
listeners. t
AUTHOR BIOGRAPHY: Bob Hertling, RCDD, OSP, is a Supervising
Engineer, Communications with the Parsons Corporation. For the past
16 years, he has supported communications and electrical design and
construction phase services on numerous intelligent transportation system
and rail/transit projects for the Massachusetts Turnpike Authority, the
Massachusetts Bay Transportation Authority (MBTA), the Port Authority of
New York and New Jersey, the Long Island Railroad, Amtrak®, New Jersey
Transit, the Southeastern Pennsylvania Transportation Authority (SEPTA)
and New York City Transit. Prior to his employment by Parsons, Bob was
also a Telephone Technician and an Electronics Engineer while on active duty
with the U.S. Coast Guard for 23 years. Bob is a 15-year member of BICSI®
and also is a member of the NFPA and IEEE®. He holds an Associate’s degree
in Telecommunications Engineering Technology and a Bachelor’s degree
in Management of Telecommunications Systems from Capitol College in
Laurel, Maryland. Additionally, he is a licensed Telecommunications Systems
Contractor in his home state of Rhode Island. He can be reached at
robert.hertling@parsons.com.

26 u ICT TODAY
The First Step
to Data Center
Design Expertise
DC101: Introduction to Data Center Design

A self-paced, online course designed

to cover foundational knowledge of
data center design.
Learn the fundamentals: data
center design process, design
criteria, systems,components,
sustainability, commissioning
and maintenance.

bicsi.org/dc101
May/June 2016 t 27
 By Josh Taylor

With the ratification of In fact, it is already an emerging influence on how

organizations plan, build and operate their existing data
new industry standards center architectures. The market proves this: manufacturers
are responding to the increased demand for Ethernet hard-
and increased demands ware, including cabling products, switches and transceivers.
on data center The implementation of 40/100 Gb Ethernet is dependent
on a variety of organizational factors, including existing
throughput, 40/100 infrastructure, budget, throughput demand and leadership
gigabit (Gb) Ethernet priority. However, it is clear that the stage is set for the most
dramatic change related to data center optical fiber infrastruc-
will be an integral tures since their inception. This article discusses the deep

component of next- impact that this network speed transition has on data center
cabling infrastructure, and the decisions that organizations
generation data centers. will need to make to accommodate these changes.

28 u ICT TODAY
Choosing the right cabling product can combat the issue of balancing manageability versus
performance. Cabling products with low optical loss rates will ensure that a structured cabling
environment is running at its peak.

Why Are Data Centers Migrating to 40/100 Gb Ethernet? particularly true for data centers built
DATA GROWTH: The world revolves around digital data. We now rely on in the 1980s, before high-performance
data to conduct business, engage in social activities and manage our lives. cabling even existed.
There is no sign of slowed growth in the production of, and demand for,
more data, or for faster access to it. According to the 2014 IDC Digital DECREASING TOLERANCE FOR DOWNTIME:
Universe Study sponsored by EMC: “Like the physical universe, the When data transactions are interrupted
digital universe is large—by 2020 containing nearly as many digital bits due to network downtime, it translates
as there are stars in the universe. It is doubling in size every two years, to a very real loss of revenue. When
and by 2020 the digital universe—the data we create and copy annually— Amazon.com® went down in August
will reach 44 zettabytes, or 44 trillion gigabytes.”1 2013, the company lost $66,240 per
minute.2 Considering how quickly lost
THE CLOUD: Among several other factors, the increase in cloud storage will revenue can add up, it makes sense that
drive the need for data throughput. “In 2013, less than 20 percent of the there is an extremely low tolerance for
data in the digital universe [was] ‘touched’ by the cloud, either stored, network downtime.
perhaps temporarily, or processed in some way. By 2020, that percentage The effect of downtime on revenue
will double to 40 percent.” 1 is even greater when considering end-
user experience. According to one
THE INTERNET OF THINGS: Another factor contributing to the exponential source, network downtime measured
growth of information is the advent of the Internet of Things (IoT). for user experience and business needs
“Fed by sensors soon to number in the trillions, working with intelligent costs an average of $5,600 per minute.3
systems in the billions, and involving millions of applications, the Network administrators should
Internet of Things will drive new consumer and business behavior that have a contingency plan in place in the
will demand increasingly intelligent industry solutions...”1 event of network failure. However, one
This exponential growth in information means processing speeds of the most effective ways to mitigate
must also increase so as not to slow access to data. High-performance this issue is to make sure the existing
cabling that can transfer data over 40/100 Gb Ethernet will be a necessary network is able to meet the demands of
addition to data centers looking to keep up with this digital data growth. increasing data throughput, including
upgrading networks capable of handling
VIRTUALIZATION: A double-edged sword, virtualization can help data 40/100 Gb speeds.
centers save on capital expenses, improve operational efficiency and
create more agile infrastructures. There are many types of virtualization, MANAGING CAPITAL EXPENSES: While
from desktop to storage to server. Server virtualization, in particular, migrating to 40/100 Gb Ethernet
calls for fewer, more efficient servers, which translates to fewer server creates an up-front capital expense, it
connections. Because there are fewer connections, however, it is saves data centers in the long run by
important that these connections work properly. Unfortunately, many future-proofing infrastructure. Not only
data centers do not contain cabling infrastructure designed to meet will data centers be prepared for the
the high-performance capabilities that virtualization demands. This is increasing demands on data throughput,

May/June 2016 t 29
 YEAR APPLICATION DATA RATE STANDARD LOSS BUDGET (dB)
1982 Ethernet 10 Mb/s* IEEE® 802.3 12.5
1991 Fast Ethernet 100 Mb/s IEEE 802.3 11
1998 Short-Wavelength Fast Ethernet 10/100 Mb/s TIA/EIA-785 4
2000 1G Ethernet 1000 Mb/s IEEE 802.3z 3.56
2004 10G Ethernet 10,000 Mb/s IEEE 802.3ba 2.6
2010 40G SR4 Ethernet 40,000 Mb/s IEEE 802.3ae 1.9
2010 100G SR10 Ethernet 100,000 Mb/s IEEE 802.3ba 1.9
2015 100G SR4 Ethernet 100,000 Mb/s IEEE 802.3bm 1.9

TABLE 1: Ethernet transmission speeds and loss amounts. *Megabits per second.

but the high-performance cabling infrastructure required of 40/100 Gb leads to decreased efficiency, increased
Ethernet can grow with future hardware upgrades. This will reduce the dB loss and more cable management
need to tear out and replace cabling with each upgrade. challenges.
The Telecommunications
Preparing for 40/100 Gb Ethernet Migration Infrastructure Standard for Data
LINK DISTANCES AND LOSS AMOUNTS: As data center speeds increase, Centers, or TIA-942, was developed
optical loss budgets decrease. Optical loss occurs over cabling distance to address various data center
and at mating points where connections are made. Since most data infrastructure design topics, including
center cabling runs are shorter distances (compared to long-haul the problem of spaghetti cabling.
campus runs), the inherent losses from distance in a data center are Among other aspects of data center
somewhat negligible compared to the losses incurred from mating planning and design, TIA-942 focuses
points. As connections in the data center increase to improve on the physical layout of cabling
manageability, performance suffers. This is because added connections infrastructure (Figure 1).
contribute to increased decibel (dB) loss. Therefore, a balance must TIA-942 offers a roadmap for
be maintained between manageability and performance. data center cabling infrastructure
Choosing the right cabling product can combat the issue of based on the concept of a structured
balancing manageability versus performance. Cabling products with cabling environment. By creating
low optical loss rates will ensure that a structured cabling environ- logical segments of connectivity, a
ment is running at its peak. When comparing dB loss rates of cabling structured cabling system can grow
products, look for “maximum” instead of “typical” loss rates. and move as data center needs
While typical loss rates can allude to the performance capabilities change and throughput demands
of a product, they are not helpful when determining loss budgets increase. Therefore, implementing a
(Table 1). structured cabling system in accordance
with the standards is the ideal way to
Cabling Infrastructure Design prepare for migration to 40/100 gigabit
Due to the exponential port growth experienced by data centers per second (Gb/s) speeds.
during the last two decades, cabling infrastructure is often reduced to The heart of a structured cabling
a cluttered tangle commonly referred to as “spaghetti cabling.” This system is the main distribution area

30 u ICT TODAY
 Offices,
Entrance Room
(Carrier Equipment & Demarcation)
Backbone Cabling Connectivity Options
Operations Center, Horizontal Cabling When migrating to 40/100 Gb
Support Rooms,
speeds, there are several connectivity
Telecom Room
options to consider when planning
Main Distribution Area Computer Room
the cabling infrastructure. The first
uses long-haul (LX) transceivers
Horizontal Distribution Area Horizontal Distribution Area Horizontal Distribution Area with singlemode (SM) cabling.
Data is transmitted via serial
Zone Distribution Area
transmission. In serial transmission,
one optical fiber is dedicated to
carry transmitting data and another
Equipment Distribution Area Equipment Distribution Area Equipment Distribution Area
carries receiving data. These two
fibers make what is referred to as a
FIGURE 1: Basic TIA-942 recommended layout. “channel.” A channel is defined as
the optical fiber, or group of fibers,
used to complete a data circuit. Until
(MDA). All equipment links back to the MDA. Other terms used to define recently, serial transmission has
this area include main cross-connect, main distribution frame (MDF) and been used for Ethernet speeds up to
central patching location (CPL). The principle of a structured cabling system 10 Gb/s.
is to avoid running cables from active port to active port (often referred to This setup is typically not used
as “point-to-point”). Instead, all active ports are connected to one area— in data centers because it is built
the MDA—where the patching is done. This is also where moves, adds and for long distances. It is also very
changes (MACs) take place. expensive despite the abundance
TIA-942 calls for the use of interconnect points, which are typically in the (and therefore low cost) of SM
form of patch panels (also referred to as fiber enclosures). Patch panels allow cabling. In order to work effectively
for patch cables (or jumpers) to be used in the front of the cabinets or racks over long distances, the lasers used
where the equipment is housed. The patch cable would then connect to an in LX transceivers are extremely
optical fiber trunk and then to another patch panel in the MDA. precise—and expensive. This
There are several advantages to implementing a structured cabling system. drastically increases the overall cost
First, using optical fiber trunks significantly reduces the amount of cabling of an LX/SM connectivity solution.
bulk both underfloor and in overhead conveyance. Implementing a structured The next option uses short-haul
cabling system also reduces airflow congestion, which reduces power usage. (SX) transceivers with multimode
Another distinct advantage to a structured cabling system is that it allows optical fiber cabling (Figure 2 on
for modularity, meaning connector changes can be made without having page 32). Data is transmitted via
to remove horizontal or distribution cabling. For example, a chassis-based parallel optic transmission. Parallel
switch with 100BFX ports is connected to a patch panel using SC optical fiber optic transmission aggregates multi-
jumpers. Upgrading the chassis and installing new blades with LC ports does ple optical fibers for transmission
not require replacing the entire channel as would a point-to-point system. and reception. For 40 Gb SR4 trans-
Instead, the module within the patch panel is replaced. Underfloor and mission, four fibers transmit at 10
overhead cabling remains undisturbed. Gb/s each, while four fibers receive
However, it should be noted that this method adds insertion loss to the at 10 Gb/s each. This means a total
channel because it adds more mating points. To offset insertion loss created of eight strands of fiber will be utiliz-
by additional mating points, high-performance optical fiber cables should be ed for a 40 Gb Ethernet channel.
used for implementation.

May/June 2016 t 31
 a significant advantage to end-users
with LC connector footprints in
their existing infrastructures.
40G SR4 100G SR10 100G SR4 The QSFP-40 Gb universal
transceiver utilizes the LC duplex
footprint but is also universal for
both MM and SM optical fiber.
This standards-based transceiver is
compliant with IEEE 802.3bm,
FIGURE 2: Parallel optic transmission over 40 and 100 Gb Ethernet. so it can inter-operate with QSFP-
40G-LR4 and QSFP-40G-LR4L.

FIBER NUMBER MAX OPERATING A newly developed bi-

SPEED LOSS BUDGET
TYPE OF FIBERS DISTANCE (METERS) MAX (DB) directional (Bi-Di) transceiver
also uses the LC duplex footprint.
10 GB OM3 2 300 2.6 The key factor for this transceiver
40 GB SR4 OM3 8 100 1.9 is its use of multiple wavelengths.
It uses two 20 Gb/s channels,
40 GB SR4 OM4 8 150 1.5
each transmitted and received
100 GB SR10 OM3 20 100 1.9 simultaneously.
100 GB SR10 OM4 20 150 1.5
Optical Fiber Types
100 GB SR4 OM3 8 70 1.8
If multimode cables are being
100 GB SR4 OM4 8 100 1.9 used to migrate to 40/100 Gb
Ethernet, OM3 or OM4 optical fiber
TABLE 2: Ethernet operating distances over optical fiber types. is required, replacing any OM1 or
OM2 fiber cables. OM4, the newest
The same principle applies for 100 Gb SR10, except the number of optical fiber type on the market,
fibers increases. Ten optical fibers at 10 Gb/s each transmit data, and 10 transmits the most bandwidth
fibers at 10 Gb/s each receive. A total of 20 fibers make up a 100 Gb SR10 and is more effective over
Ethernet channel. longer distances. OM4 is highly
With the IEEE 802.3bm standard update, a new connectivity option is recommended for any new installs
offered in 100 Gb SR4. This option allows for 100 Gb Ethernet speeds utilizing as it provides the longest lifespan
a 12-fiber multi-fiber push-on (MPO) interface. It is the same principle as in a cabling infrastructure (Table 2).
40 Gb SR4 but each optical fiber would be transmitting or receiving 25 Gb/s.
These short-haul connectivity setups are more suited for migrating to Connectivity Options
40/100 Gb Ethernet because they work well under the short distances found The LC fiber cable connector is
within a data center. Also, because SX transceivers use a vertical-cavity the most accepted connector used
surface-emitting laser (VCSEL), they are much less expensive than their in the data center, especially for
LX counterparts. high-density network applications.
The next option features new technology. Recent technological advance- The LC connector has one optical
ments have offered alternative options to the standard Quad Small Form-factor fiber and is typically mated in pairs
Pluggable (QSFP) multimode (MM) transceivers with an MPO connection. for a duplex connection (Figure 3
These new transceivers utilize a duplex LC footprint—a technology that offers on page 34).

32 u ICT TODAY
House
ad

May/June 2016 t 33
 Next Steps for Data Centers
Data centers are experiencing
the most significant change in
cabling infrastructure since the
introduction of optical fiber
cabling. No longer is it a question
of if data centers will migrate to
40/100 Gb Ethernet, but when.
Installing a high-performance
optical fiber structured cabling
infrastructure is essential to a
successful migration.
This discussion has covered
why migration to 40/100 Gb
FIGURE 3: Typical LC duplex connector and FIGURE 4: Typical MPO-style connector and
Ethernet is imminent, as well as
transceiver for 10 Gb Ethernet. transceiver for 40/100 Gb Ethernet.
the decisions data center managers
will need to make to prepare for
Possibly the most drastic change data centers will undergo in migrating implementation. There are several
to 40/100 Gb Ethernet is a change from the LC connector to the MPO-style steps that can be taken to prepare
connector (Figure 4), developed by Nippon Telegraph and Telephone Corpora- for this change:
tion (NTT). MPO is the generic term for multi-fiber push-on connectors. 1. Determine current and future
What about copper? There have been significant technology improve- data center needs, including
ments over the past few decades that create the potential for 40 Gb copper throughput demand, data
links. Choosing copper over fiber usually comes down to cost. Active copper production rates and business-
cables with transceivers on each side that utilize coaxial cables are surging in driven objectives. In what
the market, driven by top-of-rack architecture that utilizes switches at the top ways does the current data
of a rack versus a patch panel. This can be costly, especially when considering center infrastructure support
hardware refresh rates and support windows. or fail those needs?
Another option will be category 8 balanced twisted-pair cabling, which will 2. Use this information to
support 40 Gb links for channels up to 30 meters (m [100 feet (ft)]) in length. determine when the data
The standards for the media and protocols for category 8 cable are expected to center should migrate to
be published in 2016, along with a 26GBase-T alternative. This is an important 40/100 Gb Ethernet.
development that will allow smaller data centers, and any users committed 3. Map out the current data
to maintaining their copper infrastructures, to migrate to 40 Gb with fewer center infrastructure.
cost implications. This migration will be standards-based and will use the RJ45 4. Use this map to create a plan
connector for a seamless transition within the physical infrastructure. for the hardware and cabling
For the long term, it is clear that optical fiber will likely play the dominant infrastructure upgrades
role in data center structured cabling. It has better transmission properties necessary for migration.
and is not susceptible to external interference the way the copper medium is. 5. Create a plan for migration,
However, copper cabling will continue to have a role toward the edge of a including internal communica-
data center structured cabling system, as well as the edges of a campus. tion strategy, budget, timeline,
and roles and responsibilities
of those involved.

34 u ICT TODAY
 The timeline for migration is
Save the date!
different for every data center,
depending on technology needs,
budget, size and organizational
priority. However, educating the B I C S I FA L L
organization on 40/100 Gb CONFERENCE
Ethernet, evaluating current & EXHIBITION
cabling infrastructure and begin-
ning plans for implementation
will help ensure a smooth, trouble-
free migration. t

REFERENCES
1.
http://www.emc.com/leadership/digital-

House
universe/2014iview/index.htm
2.
http://www.forbes.com/sites/kellyclay/
2013/08/19/amazon-com-goes-down-
loses-66240-per-minute/
3.
http://www.datacenterknowledge.com/
archives/2011/08/10/true-costs-of-data-
center-downtime/

AUTHOR BIOGRAPHY: Josh Taylor is the

ad
senior product manager at CABLExpress,
where he is responsible for managing the
development, support, and marketing SEPTEMBER 11-15
for all product lines. Taylor has been with
CABLExpress for more than 15 years,
previously serving as an infra-structure
specialist team leader. An expert in data
center cabling trends and technologies,
Taylor produces Respect Layer One®, an
educational video series that addresses
industry standards and best practices for
data center professionals. He can be reached
at jtaylor@cablexpress.com.

Is San Antonio Ready for BICSI?

We’re about to find out! Join BICSI at the Henry B. Gonzalez Convention
Center in beautiful San Antonio, Texas, for insightful and educational
presentations, a glimpse into the latest industry technologies and multiple
networking opportunities. Registration is now available, and your company
can also become a conference exhibitor and sponsor.

Get all the details at bicsi.org/fall.

May/June 2016 t 35
 By Aaron Hesse, RCDD, PE

OPTICAL FIBER
Technology in the classroom has become
more than teaching basic computer skills.
to the Classroom
It is essential to deepening and enhancing the classroom
experience. Multimedia services in the classroom are driving
significant demand for data in higher education and K-12
enterprise networks. Between bring-your-own-device (BOYD)
policies, high-speed Wi-Fi, distance learning classrooms,
and the Internet of Things (IoT) looming on the horizon,
universities and school district network administrators and
facility managers are working to provide the speeds required
by today’s classrooms while future-proofing their network
from the coming changes. And they are increasingly bringing
fiber optics directly to the classroom.

36 u ICT TODAY
 There are two popular topologies to
choose from when deploying optical fiber
directly to the classroom: a distributed
switch network with edge switches
pushed into the classroom or a gigabit
passive optical network (GPON). These
technologies may appear similar to the end
users but there are significant differences,
and it is important to look at factors such
as initial cost of deployment, total cost
of ownership, and achievable bandwidth
prior to choosing one or the other.
THE DEMAND FOR BANDWIDTH
The modern classroom environment is
rich with technology. Professors, teachers
and school administrators are constantly
looking for new and innovative ways
to deliver educational materials to their
students in ways that are effective and
engaging. Additionally, many classroom
and building support systems such as voice
over internet protocol (VoIP), classroom
AV systems, and IP-based intercom and
paging systems are converging on the
structured cabling plant.
Many schools are implementing a
BYOD policy for use on their in-building
wireless Internet. It is not uncommon for
students to bring two or three wirelessly
connected devices to the classroom in
middle school and high school settings,
each device being capable of streaming
video, music and a torrent of social
media traffic.
Streaming video into the classroom
from a central location also is becoming
more popular. School districts find
that housing multimedia content in a
central location and accessing it from
the classroom is becoming cost effective.
This reduces the need for each teacher to
manage his or her classroom multimedia
content locally. Further, new software
suites are giving teachers the flexibility to
display their screen to the class from either
the tablet or desktop computer using the
classroom wireless network. The challenge
with these technologies is that this traffic
will be a burden to the structured cabling
plant or building wireless network.
Live streaming video is one of the
most demanding forms of network traffic.
Prerecorded streaming video, such as an
educational movie, requires an initial ax, 10 Gb/s
burst of bandwidth to fill the video buffer.
From there, maintaining the video buffer
does not require a significant amount of
bandwidth. However, live video does not
have the luxury of a large buffer. The larger
the buffer, the more lag between what is
transmitted and what is seen on the other
end. For this reason, videoconferencing
or distance learning classrooms require a
high degree of quality of service, as well
as significant bandwidth for the duration
of the transmission. If the network is
overburdened, the user may experience
delays in communication or poor video
quality, making the technology difficult to
use for its intended purpose. ac2, 3.47 Gb/s
To those who use wireless network-
ing, IEEE® 802.11a/b/g and n may be
familiar terms. In 2013 the IEEE published
802.11ac, which built on the 802.11n
standard and expanded on what is
called multiple input, multiple output
(MIMO) technology; the first devices ac1, 1.3 Gb/s
to use this standard could achieve one
gigabit (Gb) speeds. Starting in 2014 and
2015, manufacturers began producing
n, 600 Mb/s
what are called 802.11ac Wave 2 devices.
g, 22 Mb/s
Speeds from these access points can reach b, 15 Mb/s
6.8 gigabits per second (Gb/s). In 2015, FIGURE 1:
manufacturers began offering what are Comparison
of Theoretical
being called multigigabit Ethernet ports
Bandwidths of
with some of their Ethernet switches. IEEE 802.11
These ports offer 2.5 and 5 Gb/s speeds Standards.
May/June 2016 t 37
 MDF/Building Distributer
Internet
Layer 3 Core Switch
Fiber Switch Stack
Fiber Patch Panel
FIGURE 2: Example of a Distributed Switch Topology in a School.
over a single legacy category 5e
or 6 cable. IEEE 802.11ax, the
current wireless standard under
development, is predicted to have
top speeds of around 10 Gb/s
(Figure 1 on page 37).
DISTRIBUTED SWITCH
The first option for delivering
fiber to the classroom is with a
distributed switch topology that
places edge switches in the classroom.
This approach uses a hierarchical
star configuration and pushes a
telecommunications enclosure out
to the classroom. The layout of
the system closely matches a more
traditional network configuration.
Each classroom has an 8- to 16-
port switch and one or two 10 Gb
multimode optical fiber uplinks to
the floor distributor or the building
distributor. Some switches can also
provide two multigigabit Ethernet
ports, supporting future high-speed
wireless networking capabilities.
38 u ICT TODAY
Category 6
t Students
Multimode Fiber
In this topology, the building’s
main distribution frame (MDF)
requires the use of a Layer 3
optical fiber switch stack for switch
aggregation. A fixed aggregation
switch can provide numerous 10 Gb
small form-factor pluggable (SFP+)
ports for use with the multimode
optical fiber to the distribution
switches in the classrooms. If the
building is a single story, this
may be the only dedicated telecom-
munications space (Figure 2). If the
building has multiple stories, ANSI/
TIA-4966 Standard for Educational
Facilities requires the use of one floor
distributor telecommunications
space per floor. Note that a network
enclosure does not satisfy the
requirements of a floor distributor.
The optical fiber connection
from the telecommunications room
(TR) to the classroom should be rated
for use with 10 Gb systems at this
distance. Although BICSI’s current
standard on the subject, ANSI/
Classroom
Wireless
Access Point
t
t Teachers
Telecommunications
Distribution Enclosure
Switches
BICSI 001-2009 Information Transport
Systems Design Standard for K-12
Educational Facilities, states that OM1
multimode fiber be the minimum
requirement, the upcoming rewrite
of that standard states OM3 at a
minimum with a recommendation
for OM4. The rewrite for ANSI/BICSI
001-2009 is currently titled BICSI
D003 Information and Communication
Technology Systems Design and Imple-
mentation Best Practices for Educational
Institutions and Facilities and is in the
approval process.
According to BICSI’s 13th
edition of the Telecommunications
Distribution Methods Manual (TDMM),
OM3 multimode fiber is rated for 10
Gb speeds up to 300 meters (m) and
OM4 is rated for up to 550 m. Using
the distributed switch topology, it is
important to verify optical fiber run
distance, as these runs can get fairly
long in large, multistory buildings.
While designing these systems, keep
in mind that there may be a 3 m
 GPON is a particularly attractive option in messy remodels
Technology
Infrastructure
32% or retrofit projects. In many cases, the building structure
will not support the additional required size of a category 6
Active Equipment
+29 cabling plant. One question that will need to be answered
is: will 1 Gb to the classroom be enough? Using the GPON
Building IT Spaces
-39 solution, that single optical fiber to the classroom can be
split to feed multiple ONTs.
TOTAL SAVINGS
3%
TABLE 1: Cost of distributed switch network
versus traditional central switched network.

[10 feet (ft)]) to 4.6 m (15 ft) service
loop that will need to be brought
into consideration.
MIDDLE SCHOOL EXAMPLE
The cost difference between a
distributed switch network and a
traditional central switched network
was analyzed using a two-story,
70,000-square-foot middle school
as an example (Table 1). Costs were
calculated using quoted pricing
for equipment and RSMeans 2015
adjusted for the state of Washington.
Analysis of capital expenditures
(CapEx): Technology infrastructure
is reduced due to the lower cost
of multimode optical fiber to each
distribution switch compared to
a multiple category 6 structured
cabling path. Pathway costs are also
reduced due to the elimination of
large cable trays down hallways and
a significant reduction in required
vertical pathways.
The cost of active equipment
would increase when using a
distributed switch topology. The
required number of fiber switches
and distribution switches would
increase. Additionally, a distributed
switch network requires either a
central uninterruptible power supply
(UPS) with distributed power or
numerous smaller UPSs for each
intermediate distribution frame
(IDF) in the classroom.
There is some additional cost
associated with the telecommunica-
tions enclosure and conditioning
required to enclose the distributed
IDFs. In the model above, it was
assumed that these were going
to be stand-alone enclosures,
although some districts choose to
provide space in the cabinetry if
the network overhaul is part of an
overall building remodel. This could
increase the savings to the project at
the cost of usable cabinet space.
SYSTEM PERFORMANCE
AND FUTURE-PROOFING
The savings to the initial
construction project are marginal.
However, while this option may not
be the best choice for those looking
to save project costs, the additional
benefits come in performance and
future-proofing. With an unprece-
dented 10 Gb to the classroom, this
topology will support the needs
of the instructor well beyond the
expected life cycle of the network.
Another advantage to this
system is the availability of
multigigabit ports in smaller
switches. This allows for six
10/100/1000 Gb Ethernet ports,
sufficient for most classrooms,
plus two 100/2500/5000/10,000
multigigabit ports to be used with
the wireless access points (WAPs).
This will support IEEE 802.11ac
Wave 2, as well as 802.11ax, without
any need for switch or cabling
upgrades.

May/June 2016 t 39
 MDF/Building Distributer
Internet
Layer 3
Core Switch
MPO Cable
Optical Line Terminal (OLT)
GIGABIT PASSIVE OPTICAL
NETWORK (GPON)
Passive optical networks are not
new to the landscape. Numerous
government bodies have begun
standardizing around GPON
technology; this includes all four
branches of the U.S. military and the
Departments of Defense, Homeland
Security and Energy. GPON also has
been deployed in large commercial
buildings.
Similar to the distributed switch
model, the network requires a layer
3 switch and an incoming WAN or
Internet connection. From there the
design deviates considerably. The
installation requires the use of an
optical line terminal (OLT) in the TR.
This replaces the layer 3 aggregation
switch stack found in a typical
building distribution room. This unit
comes in many different sizes and is
generally modular in construction
40 u ICT TODAY
IDF/Floor Distributer
Singlemode Singlemode
Fiber Fiber
t 1:16 t
Splitter
Optical Network Terminal
Fiber Hub
FIGURE 3: Example GPON Topology in a School.
and easily scaled for future upgrades.
From this unit, singlemode optical
fiber is routed through passive
splitters to either multi-fiber push
on (MPO) cables or multiple single-
strand singlemode fibers. These
fiber cables are then distributed
throughout the building before
reaching the classroom (Figure 3).
An optical network terminal
(ONT) is required to convert the
singlemode optical fiber to category
cabling, but this can come in either
a multiport rack-mounted switch
housed in the millwork, a desktop
unit that serves the workstation, or a
recessed wall unit that emulates the
look of a typical two-port data outlet
and fits in a single gang box. Figure
3 uses a rack-mounted switch in
order to remain as similar as possible
to the distributed switch topology.
GPON is a particularly attractive
option in messy remodels or retrofit
Classroom
Wireless
Access Point
t
(ONT) t Teachers
t Students
projects. In many cases, the build-
ing structure will not support the
additional required size of a
category 6 cabling plant. Perhaps
the existing data plant is category 3,
and the concrete construction of
the building would require coring
out every existing pathway in order
to upgrade to a gigabit network.
In cases like this, the entire copper
cabling plant can be replaced with
one or two singlemode optical fiber
cables. This greatly reduces the
cost of pathway and can present
significant cost savings.
UNIVERSITY BUILDING
EXAMPLE
Universities are pioneering the
use of passive optical networks. All
eyes seem to be on Washington State
University in the Pacific Northwest.
Dozens of universities call them to
learn more about the technology,
 and many have traveled to visit With the total construction cost

-50%
Technology
them personally. for GPON at nearly half the cost of
Infrastructure To illustrate the cost savings of a traditional switched network, the
their installations, the university total savings to the construction

-28
has studied two of their recent project was found to be over half a
Active Equipment
building remodels that occurred a million dollars.
couple of years from each other,
in two buildings that are nearly Analysis of CapEx: This data

-70
Building IT Spaces identical. One has been remodeled reveals a number of key differences.
with a traditional switched network First, these buildings are five stories
and the other with GPON. The high. In a smaller building, or a

41%
figures in Table 2 reflect the actual
TOTAL SAVINGS
CapEx incurred by the university,
not a budgetary estimate prior to
construction.
K-12 building with a large footprint
but only one story, the savings
associated with a reduction in
telecommunications space would

TABLE 2: Cost of GPON installation

over traditional switched network.

Ignite the Cloud!

LaserWave® FLEX WideBand Optical Fiber
Enabling low cost multimode solutions for next generation data center networks

To learn more, visit www.ofsoptics.com

/ofsoptics /ofs_telecom /company/ofs /OFSoptics

May/June 2016 t 41

be less significant; i.e., a building that may
have only had two TRs to begin with would
only reduce their telecommunications
spaces by one room instead of four.
This cost breakdown does not
reflect the additional monetary value
of the space that is saved by reducing
telecommunications space. That benefit
will have to be calculated based on the
building’s application. If the building is a
dorm room or apartment, and the space
is revenue generating, the calculation
becomes fairly straightforward. The
monetary value of the space saved can
be seen as a direct result of their network
topology decision. For university or school
districts, the removal of a TR may result in
additional storage or support space, a much
more difficult value to quantify.
Another aspect of these numbers is
that this represents a single building.
If there are multiple buildings on the
property, this cost breakdown would apply
to the first building only. The cost benefit
to the subsequent remodels to buildings
on the property would be substantial.
While the first building may save 30-40
percent over a traditional installation, the
cost savings to the next building could
be as much as 60 percent. As long as the
original OLT can support the additional
ONTs of the nearby building, the additional
buildings would not require an OLT or a
dedicated space for an MDF.
SYSTEM PERFORMANCE AND
FUTURE-PROOFING
One question that will need to be
answered is: will 1 Gb to the classroom be
enough? Using the GPON solution, that
single optical fiber to the classroom can
be split to feed multiple ONTs. This would
provide 1 Gb to the desktop if necessary.
However, if WAPs are expected to utilize
the IEEE 802.11ac Wave 2 or 802.11ax
42 u ICT TODAY
ADD THIS GLOBAL STANDARD TO YOUR REFERENCE LIBRARY!
performance capabilities, the technology
infrastructure will fall short of delivering
these speeds.
Furthermore, the upcoming XPON or
10GPON standard, needed to provide 10
Gb/s to the ONT, is not currently defined.
Predictions suggest that the OLT line cards
within the chassis and all ONTs will need
to be replaced in order to take advantage of
the 10GPON standards. If this is expected
to be required during the lifespan of the
installation, this could be a costly upgrade.
CONCLUSION
Universities, school districts and their
consultants need to analyze the goals
of the network installation and select
a topology and technology that makes
sense for the end user. There are numerous
options available and it is important to
explore each option. If optical fiber to
the classroom is a priority, each approach
discussed above has its own strengths.
Whether the decision is driven by
economics or by the need for significant
bandwidth in the future, bringing optical
fiber directly to the classroom might be a
good fit for numerous school districts. In
the face of a changing data consumption
landscape and constrained budgets,
network administrators are often required
to choose between network performance
and more economic options. For those
looking for ways to provide a 21st century
network while saving and conserving
funds, GPON is an attractive technology.
For those with a focus on future-proofing
and providing significant network speeds,
consider looking at a distributed switch
technology in order to provide optical fiber
to the classroom. t
AUTHOR BIOGRAPHY: Aaron Hesse, RCDD, PE, is
an electrical engineer with Coffman Engineers in Spokane,
WA. He is a licensed professional engineer and a BICSI
Registered Communication Distribution Designer.
He can be reached at hesse@coffman.com.
The Essential Standard
for Data Center Design
an American National Standard

ANSI/BICSI
002-2014
Data Center Design and
Implementation Best Practices VITAL DATA CENTER
®

DESIGN INFORMATION

Get your copy of ANSI/BICSI 002-2014,

Data Center Design and Implementation
Best Practices (Available in English and Spanish)
A comprehensive document covering:
n Modular and “container” data centers

n Data center infrastructure management

(DCIM)
n DC power

n Hot and cold aisles

n Multi-data center architecture

n Data center service outsourcing

n Energy efficiency

n Objective methodology to determine

resilience requirement
n And more

Learn the specifics of the planning, design and

installation of data centers from a respected and
trusted source in the information and communications
technology (ICT) industry.

Buy it today at bicsi.org/002.

May/June 2016 t 43
 DisplayPort and the
There was a time when it was easy to understand AV connectivity.
Consumer products used an RCA connector
(one for composite, three for component, and we
won’t mention S-Video) and commercial products
used BNC connectors (one for composite, five for
RGBHV [for red, green, blue, horizontal sync, vert-
ical sync] and again, we won’t mention S-Video).
That pretty much summed it up for decades, with
one glaring exception: the computer. Both laptop
and desktop computers needed something more
compact and with some advanced capabilities—
such as the ability to exchange extended display
identification data (EDID) information with the
display. A different connector was needed.
44 u ICT TODAY
By Joseph D. Cornwall,
CTS-D, CTS-I
of Desktop
Video
Connectivity
Solutions
Connectivity Technologies Develop
Over the Years
The video graphics array (VGA) was first
introduced with the IBM® PS/2 line of computers
in 1987 (VGA also refers to a specific resolution,
but this discussion focuses on the cables and
connections). The VGA connector uses a three-
row, 15-pin DE-15 that carries analog component
RGBHV video signals, and Video Electronics
Standards Association (VESA) display data channel
(DDC, also called EDID) data. For nearly three
decades, this blocky analog connection ruled
the office. While not particularly elegant,
it worked and was well understood.
Low-voltage differential signaling
(LVDS) is a physical layer specification
that describes the way signals move
between inputs and outputs on integrated
circuit (IC) chips. In use since 1994,
LVDS was used for computer video and
graphics data transfers and drove the most
advanced versions of the VGA interface.
LVDS became an increasingly important
technology as computer resolutions
began to escalate beyond 800 x 600
super VGA (SVGA), but the proliferation
of high-definition digital video at the
consumer level changed that. This is
because widespread proliferation of HDTV
came with an unanticipated challenge.
Intellectual rights management for content
is a particularly thorny issue in the world
of digital media. Unlike analog, there is no
generational loss of quality when copies
are made. In the digital world, there are
no copies—there are only clones.
Often erroneously referred to as “high
definition copyright protection,” high
bandwidth digital content protection
(HDCP) is a system developed by Intel®
Corporation. HDCP leverages a key
exchange protocol known as Blom’s
Scheme. It is intended to prevent
encrypted content from being played on
unauthorized devices or devices which
have been modified to copy HDCP content.
Before sending data, a transmitting device
checks that the receiver is authorized to
receive it. If so, the transmitter encrypts the
data to prevent eavesdropping as it flows
to the receiver. As computers and personal
digital portables became a bigger part of
the industry, the ability of these devices
to interface with just about everything
became an important market factor.
On December 8, 2010, a joint
press release from AMD®, Dell®, Intel
Corporation®, Lenovo®, Samsung®
Electronics LCD Business, and LG® Display
announced “intentions to accelerate
adoption of scalable and lower power
digital interfaces such as DisplayPort (DP) Like HDMI,
and high-definition multimedia interface
(HDMI) into the PC” as an alternative to
DisplayPort
the aging, analog VGA connector. The supports fully
press release went on to say that “Intel
plans to end support of LVDS in 2013 and
embedded
VGA in 2015 in its PC client processors and digital audio for
chipsets.” The death notice for the VGA
connector had been posted. surround sound
applications
The Rise of DisplayPort
DP is a digital display interface stand- (typically eight
ard developed by VESA. It is specifically
designed for the transfer of video, audio
channels)
and data between a source and sink. and is fully
DP has replaced LVDS on essentially all
computers because of its rich feature set
compliant with
and its compatibility with transition- high bandwidth
minimized differential signaling (TMDS)-
encrypted HDMI and HDCP technologies. digital content
DP has been included in about one-in-20
protection
commercial desktops and one-in-50
laptops as of 2009. Today, DP is found on (HDCP)
the majority of new convertible, laptop,
desktop and workstation computers. Where
for content
we used to see a VGA connector, today we protection
see DP or mini DP.
Although DP has a lot of the same
and system
functionality as HDMI, it is a complemen- integration.
tary connection and not necessarily a
competitive one. This is a good thing,
because HDMI and DP deliver very
different strengths, which translate to
unique features and benefits best aligned
with their respective applications. HDMI
and DP, despite superficial similarities,
serve different markets, and there is plenty
of room for them to coexist in the AV
ecosystem.
The fundamental difference between
DP and HDMI is the use of TMDS in the
HDMI format, and the use of serial data
May/June 2016 t 45

FIGURE 1: HDMI uses TMDS, based on IBM’s 8b/10b line code.
transmission in DP. HDMI uses TMDS based
on IBM’s 8b/10b line code, which maps 8-bit
symbols to 10-bit symbols to achieve DC-
balance, bounded disparity and clock recovery
(Figure 1).
The HDMI connector consists of 19 pins,
12 of which carry red, green, blue (RGB) and
clock data across four shielded twisted pairs.
The remaining pins negotiate hardware
handshake and voltage assertion, provide
DDC EDID connectivity and support optional
connectivity unique to the HDMI (think
multimedia) application.
The DP system uses packetized data, a form
of digital communication familiar from its use
in Ethernet, peripheral component interconnect
(PCI) express, and USB technologies. The DP
protocol is based on micropackets, which embed
the video sync (clock signal) into the data
stream, allowing for a complete video link with
embedded audio (eight channels) over a single
lane (Figure 2).
The DP system features four such lanes
in its 20-pin connector topology. Each lane
occupies three pins and connects to a shielded
twisted pair of the cable. Each lane can deliver
all the information necessary to support a full
1080p video stream; this is the real power of
DP. One physical connector can deliver up
to four discrete channels of communication,
supporting the multi-monitor desktop that is
now customary. This is also known as multi-
stream transport (MST).
46 u ICT TODAY
There are several variations of DP, including Thunderbolt,
in use today. Of critical concern are the differences between
DisplayPort 1.1 (DP) and DisplayPort 1.2 (DP++), also known
as “dual mode” DP. There are no differences in the cable
and connector topology per se, but there is one important
difference in performance. DP++ has the ability to leverage
the power of the host computer to output a signal in a true
TMDS HDMI-compliant format. That means that devices that
support DP++ (and are so marked) can use a passive cable with
a DP connector on one end and an HDMI connector on the
other to make the conversion passively.
When a DP++ source is connected to the HDMI input on
a display, it can “see” that there is a demand for a clock signal
and, therefore, recognize that the signal is not packetized.
Using all four lanes of connectivity, the DP++ video card takes
a single content stream and repackages it into RGBs, sending
the signals over the four lanes of the connection. In this way,
it moves from operating in a world of four independent serial
streams to one in which a unified signal is sent over four lanes
simultaneously.
DP devices require an active device to make this
conversion. DP can only output a monolithic AV signal that
occupies a single lane, so its serial content must be actively
reformatted into a TMDS RGB state. This demands an active
device, external to the source. Since the newly released USB
Type-C technology that will make such a huge impact in the
mobile market is design limited to DP, keeping this critical
difference in mind will play an important role in designing
AV connectivity systems.
The newest version is DP 1.3. Released in 2014, it will
become more important as devices move from high definition
to ultra-high definition (UHD) 4K performance levels. At this
point in time, the distinction is not critical when selecting
and accommodating desktop connectivity solutions as the
differences are not in the physical layer.
Like HDMI, DP supports fully embedded digital audio for
surround sound applications (typically eight channels) and is
fully compliant with HDCP for content protection and system
integration. An even more important feature is the ability of
DP to support multi-monitor MST where a single output on
the computer can deliver as many as four discrete video feeds.
For a number of technical reasons, DP connectivity
is limited to a maximum run length of ≈10 meters (m [33
feet (ft)]) across native DP cables. Beyond this, the signal

FIGURE 2: The DP protocol is based on micropackets, which embed the
video sync into the data stream, allowing for a complete video link with
embedded audio (eight channels) over a single lane.
must be converted if longer runs are needed. Conversion
can be accomplished by using modems with fiber links
(maintaining the DP packetized format) for runs of tens
of meters to a kilometer or more. DP can also be adapted
to a category UTP environment for runs up to 100 m (328
ft) by using HDBaseT technology. The following is a closer
look at some of the DP solutions available for individual
installations.
DisplayPort Cables
Stock DP cables come in lengths from ≈1–10.7 m
(3–35 ft). There is no difference between DP 1.1 and
1.2 cables, and any certified cable can be used for either
application.
DP multimode cables in lengths up to 3 m (10 ft) may
support 4K UltraHD content capability. This is a useful
length for desktop connectivity or a jump into a format
converter such as an HDBaseT transmitter or fiber modem.
DP multimode interconnects of 5 m (16.5 ft) to 10 m
(33 ft) will support 32-bit payloads up to 1920 x 1200.
Although it looks quite different, mini DP is just a
different form factor and has identical properties to full
sized multimode DP links. Mini DP is often used on laptop
and tablet computers because of its smaller size and profile.
Adapters are available for mini DP to full sized DP. Cables
with mini DP on one end and full sized DP on the other
are also available in lengths less than ≈3 m (10 ft), as are
mini DP to mini DP patch cords.
DisplayPort Extenders
When the project requires links greater than ≈10 m
(33 ft), active solutions must be employed. HDBaseT is
an IEEE® standard technology that can extend digital
video signals up to 100 m (328 ft) by using transmit and
receive electronics with a category cable as the link in-
between. This also gives the option of running DP in a
plenum environment by using communications plenum
(CMP) or FT6 rated category 6 or 5e. For best results, use
solid-core screened unshielded twisted-pair category 6
cable for all digital AV applications.
Depending on the project, HDBaseT solutions can be
selected to provide a DP-to-DP configuration or a DP-to-
HDMI configuration. Performance of the two is identical
and this technology is independent of DP++ capability,
so it works with any DP device.
DisplayPort Dongles and Converters
Dongles that convert DP to analog VGA, digital
DVI-D and HDMI are all readily available, often
in a choice of black or white to match the desktop
environment. Particular attention must be paid to
systems where conversion to VGA or other analog
interfaces is necessary. Keep in mind that such a
conversion does not eliminate the need to comply
with HDCP requirements. HDCP-encrypted content
may, under no circumstances, be converted to
any unprotected format; HDCP continuity must
be maintained for an image to be displayed. Also,
conversion to the analog domain (VGA or composite
video) requires digital-to-analog conversion, an active
process that must be powered by the signal bus or
external power source. t
AUTHOR BIOGRAPHY: Joseph D. Cornwall, CTS-D, CTS-I, is a
Technology Evangelist with Legrand North America. Cornwall has held
both management and technical positions in broadcast, residential, and
commercial market sectors. Honored as the 2014 InfoComm Educator of the
Year, he is widely recognized as an energetic and compelling presenter, trusted
subject matter expert and seasoned industry professional. Cornwall regularly
addresses groups both large and small on topics as diverse as AV technology
and system design, technology trends, sales skills development and market
navigation. He can be reached at joseph.cornwall@legrand.com.
May/June 2016 t 47
 THE UNEXPECTED IMPACT OF
RAISING
DATA CENTER
TEMPERATURES
By Wendy Torell, Kevin Brown and Victor Avelar
LOW
A revision to ASHRAE standard TC9.9
released in 2011 has initiated an ongoing
industry effort to raise IT operating
temperatures. ASHRAE’s recommendation
is intended to lower data center energy
consumption by expanding the number of
hours of what is called “free cooling.”
Despite this industry effort, many data
centers still operate at temperatures at or
below 21 degrees Celsius (°C [70 degrees
Fahrenheit (°F)]). While there is an arguably
conservative culture in the industry, the
question remains: why are so few data
48 u ICT TODAY
HIGH
centers raising IT inlet temperatures? If
there were great savings to be had, one
would think that more data centers would
take advantage of these savings even given
the conservative nature of the industry.
Research has determined that data
center managers generally ask two
questions that prevent them from raising
their data center inlet temperature:
u How much energy can be saved by
increasing IT temperature?
u Will raising temperatures impact the
reliability of IT gear?
To answer these questions, it is impor-
tant to look at the data center holistically,
as the system dynamics are complex; the
energy consumption of some systems
decrease while others increase. Consider
a packaged chiller design. When the IT
temperature set point is increased, the
chiller energy decreases, for two reasons:

WHEN OPERATORS THINK ABOUT RAISING TEMPERATURES IN
THE DATA CENTER, IT IS COMMONLY UNDERSTOOD TO MEAN
RAISING THE TEMPERATURE TO A NEW FIXED SET POINT.
the data center can operate in economizer mode(s) for a larger
portion of the year, and the chiller efficiency increases. But this is
not the entire picture. Although the chiller energy decreases, the
following also occurs:
u The dry cooler (which operates in economizer mode instead of
the chiller) energy increases because the number of economizer
hours increases.
u Server energy increases because requirements for airflow,
measured in cubic feet per minute (CFM), increase as temp-
erature rises (See sidebar, “Why do server fans ramp up?”).
Computer room air handler (CRAH) fans speed up to support
u
the higher server CFM requirement, which means greater
u
CRAH energy consumption.
If not already oversized to accommodate the additional airflow,
more CRAHs are needed to match the higher server fan CFM
requirements. This means additional capital expense.
Figure 1 on page 50 illustrates these countering effects. This
article walks through an analysis of a data center with a packaged
chiller cooling architecture, to demonstrate how location and server
fan behavior have a significant influence on the potential savings (or
cost penalties) when IT inlet air temperature set points are increased.
The research also considered the implications of fixing the temp-
erature (at a higher point) versus allowing the data center tempera-
tures to float within a defined range, as the outdoor temperature
fluctuates. The final consideration was a scenario where an existing
data center is oversized (50 percent) to illustrate the impact that
percent load has on these results.
ANALYSIS OF THE DATA CENTER
The impact of raising IT temperatures on energy consumption can
vary significantly depending on the cooling architecture, the climate,
IT fan speed and percent IT load. This article looks at one architecture
and models it in varying climates to demonstrate the complex nature
of data centers, and to illustrate the importance of understanding the
risks and rewards before operational changes are made.
WHY DO
SERVER FANS
RAMP UP?
The purpose of server fans is to cool the
components inside the server chassis. The
most important of these components is the
CPU chips which can reach temperatures
upwards of 90 °C (194 °F). As the IT inlet
air temperature increases, so will the CPU
temperature. This typically triggers server fans
to increase airflow in an effort to reduce the
CPU temperature. This increase in airflow
consequentially increases server energy
consumption.
OPERATING
TEMPERATURES
OF CHILLERS
Every chiller has a maximum chilled water
temperature it is capable of supplying. This
is limited by the type and design of the
chiller. For example, in centrifugal chillers the
compressor must be capable of reducing its
speed to produce lower refrigerant pressures
without damaging the motor or without
leaking its lubricating oil into the refrigeration
circuit. Depending on the chiller type, other
chiller components may require special
features which allow for higher chilled water
temperatures. Consult with the chiller vendor
before increasing the chilled water set point.
With a high-efficiency chiller located in a mild
climate, the chiller energy savings gained by
increasing IT inlet temperatures may not be
enough to justify the increase in energy from
other devices in the data center (e.g. CRAH,
dry cooler, cooling tower, IT).
May/June 2016 t 49
 FIGURE 1: System dynamics are complex, so there is a need to evaluate the data center holistically.
ARCHITECTURE ANALYZED
This analysis looks at what is believed to be a very
common cooling architecture deployed in data centers
today—a packaged air-cooled chiller with economizer
(Figure 2). The dry cooler, utilized during economizer
mode, is a heat exchanger that directly cools the data
center chilled water when the outside air conditions are
within specified set points. Pumps move the chilled water
through the dry cooler where the cold outside air cools
the chilled water that supplies the CRAHs.
The main assumptions used in the analysis are:
u 1 megawatt (MW) data center, fully loaded
u Three air-cooled chillers in an N+1 configuration,
sized for 20-year extreme temperature
u All chillers (including the redundant chiller) operate
at part load under normal operation
u Chillers are capable of operating at higher chilled
water temperatures (see sidebar, “Operating
temperatures of chillers,” page 45)
u Use of a variable frequency drive (VFD) dry cooler
for economizer mode (no evaporative cooling)
u Fixed-speed pumps
u CRAHs with hot aisle containment in an N
configuration
Indoor CRAH
u u
u
Air cooled chiller
FIGURE 2: Diagram of the packaged air-cooled chiller architecture analyzed.
50 u ICT TODAY
u Airflow demand of servers matched with CRAH
airflow supply (i.e., CFM of servers = CFM of
CRAH fans)
u Power density of 4 kilowatts (kW)/rack.
u 3 percent cost of capital used for total cost of
ownership (TCO) calculations
u $0.10 per kilowatt hour cost of electricity
u Weather bin data from ASHRAE Weather Data
Viewer 5.0
Three different operating temperature scenarios were
created in order to compare the energy consumption
and TCO of each:
1. The baseline case assumed a fixed IT inlet tempera-
ture of 20 °C (68 °F), which is a typical operating
point for data centers today.
2. The second case allowed temperatures to float from
15.6 °C (60 °F) to 26.7 °C (80 °F).
3. The third case fixed the temperature at 26.7 °C (80 °F).
The data center scenario was analyzed in three U.S.
cities—Chicago, Seattle, and Miami—to illustrate the
impact of varying climates on the results.
ANALYSIS METHODOLOGY
Energy cost and capital expense of the entire
cooling system were analyzed utilizing the following
Dry cooler methodology:
Bin data from ASHRAE Weather Data Viewer 5.0
were used to calculate the cooling system energy at
every 1.11 °C (2 °F) bin using thermodynamic
principles and formulas. Inputs to this model
included CRAH coil effectiveness, average delta T
across IT equipment, and equipment losses for IT
fans, chillers, dry cooler, pumps and CRAHs.
 Chicago Fixed $1,186,000 $162,000 $0 $571,000 $258,000 1% Energy Reduction
Chicago Floating $855,000 $241,000 $173,000 $583,000 $320,000 No change in TCO

Seattle Fixed $1,196,000 $313,000 $0 $571,000 $258,000 15% Energy Reduction

Seattle Foating $565,000 $423,000 $175,000 $583,000 $321,000 6% better TCO

Miami Fixed $2,420,000 $13,000 $0 $571,000 $258,000 11% Energy Increase

Miami Floating $2,224,000 $54,000 $342,000 $583,000 $386,000 6% worse TCO

FIGURE 3: Summary of results from baseline of 20 °C (68 °F) fixed to floating from 15.6 °C (60 °F) to 26.7 °C (80 °F) at full load.

u The 20-year extreme temperature was used as the
worst case outdoor temperature for sizing the
packaged chiller. This design point is the generally
accepted practice for sizing chillers and is
recommended by the Uptime Institute.1
u The cooling system energy is dependent on the
different operating modes: full mechanical cooling,
partial economizer mode, and full economizer mode.
The number of hours spent in each operating mode
was calculated.2
u The IT inlet air set point was used to calculate the
chilled water temperature. The chilled water temp-
erature was allowed to range from 7.3 °C (45 °F) to
32 °C (90 °F).
u For IT inlet temperatures above 20 °C (68 °F), the
increase in server energy consumption was added to
the total cooling system energy consumption.
u The floating temperature scenario represented an
ideal case where the chiller and economizer controls
allow chilled water temperatures to reset dynamically.
In most data centers, the chilled water temperature
is set at a fixed temperature year-round and would
yield lower energy savings than this model projects.
u The capital expense values were estimated using
component, labor and design prices typically seen
in a 1 MW data center project. The change in CRAH
capital expense as the IT CFM changes with IT inlet
temperature was also accounted for.
FINDINGS
First, the findings of the baseline [where IT tempera-
tures are fixed at 20 °C (68 °F)] were compared to the
second case (where IT temperatures float up and down).
Following these findings, the comparison of the baseline
to the third case (where IT temperatures are fixed at a
higher temperature of 26.7 °C [80 °F]) is presented.
BASELINE VERSUS
FLOATING TEMPERATURES
The TCO differences of the baseline versus the float-
ing temperature case are presented in Figure 3. The TCO
shown excludes the capital cost of systems that do not
change between the two scenarios. This analysis leads to
the following conclusions:
u While chiller energy always improves (decreases),
the net energy consumed does not always improve.

1
Uptime Institute’s “Data Center Site Infrastructure Tier Standard: Topology,” http://www.gpxglobal.net/wp-content/uploads/2012/10/TIERSTANDARD_Topology_120801.pdf
2
In full economizer mode operation, the outdoor conditions allow for all mechanical cooling (i.e. those components used in the refrigeration cycle) to be turned off to conserve energy while still
effectively cooling the defined load. When the outdoor temperature limits full economizer mode operation, the cooling plant enters a partial economizer mode of operation, where a proportion
of the cooling is handled by the economizer mode and the remaining is handled by the mechanical system. The proportion of each changes (increasing the mechanical cooling proportion as
temperature increases outdoors) until full mechanical system operation is required.

May/June 2016 t 51
 CHICAGO SEATTLE MIAMI

Total energy (kWh) 0.98% savings 14.51% savings 11.01% increase

pPUE (cooling only) Improves from 1.203 to 1.178 Improves from 1.222 to 1.166 Improves from 1.327 to 1.312

Floating temperature range 15.6–23 °C 15.6–26.7 °C 15.6–21 °C

with lowest TCO (60–74 °F) (60–80 °F) (60–70 °F)

TABLE 1: Summary of results from baseline of 20 °C (68 °F) fixed to floating from 15.6 °C (60 °F) to 26.7 °C (80 °F) at full load.

FIGURE 4: Summary of results from

baseline 20 °C (68 °F) fixed to varied
max floating from 15.6 °C (60 °F) to
26.7 °C (80 °F) at full load.

u Higher IT inlet temperatures cause an increase in
IT equipment airflow which decreases the difference
in temperature (deltaT) across the CRAHs. More
CRAH airflow is needed to remove the same amount
of heat at these lower deltaT values.
u The required CRAH capacity increases at higher
chilled water supply temperatures because the heat
removal capacity of the coil decreases as deltaT
decreases.
u The degree to which the increase in energy occurs
for the servers and CRAHs depends on the IT
equipment characteristics. This is explained in the
following section.
u Bin weather data is a significant driver in determin-
ing whether floating temperatures from 15.6 °C
(60 °F) to 26.7 °C (80 °F) result in a cost savings.
Table 1 provides additional results, including
differences in the total energy kilowatt hour (kWh) and
partial power usage effectiveness (pPUE).3 Although power
usage effectiveness (PUE) improved in all cases, energy
did not always improve. This points out the limitation of
using only PUE as a basis for operational decisions.
Also highlighted are the maximum float temperature
that would result in the lowest TCO for each of the three
cities. As the data demonstrates, this optimal temperature
varies quite a bit from one city to the next.
Figure 4 is a graph illustrating the TCO as the maxi-
mum float temperature was varied. In all cases, the mini-
mum float temperature was assumed to be 15.6 ˚C (60 ˚F).
This graph demonstrates how bin data can have
a significant impact on results. In Seattle, the optimal
temperature for this cooling architecture occurs at

3
In this analysis, pPUE represents only the cooling system losses.

52 u ICT TODAY
 Chicago 20C $1,186,000 $162,000 $0 $571,000 $258,000 13% Energy Increase
Chicago 27C $855,000 $241,000 $345,000 $583,000 $388,000 5% worse TCO

Seattle 20C $1,196,000 $313,000 $0 $571,000 $258,000 2% Energy Reduction

Seattle 27C $565,000 $423,000 $345,000 $583,000 $388,000 1% better TCO

Miami 20C $2,420,000 $13,000 $0 $571,000 $258,000 11% Energy Increase

Miami 27C $2,224,000 $54,000 $345,000 $583,000 $388,000 6% worse TCO

FIGURE 5: Summary of results from baseline of 20 °C (68 °F) fixed to a higher fixed temperature of 27 °C (80 °F) at full load.

CHICAGO SEATTLE MIAMI

Total energy (kWh) 12.52% increase 2.34% savings 11.16% increase

pPUE (cooling only) Improves from 1.221 to 1.182 Improves from 1.241 to 1.170 Improves from 1.363 to 1.312

TABLE 2: Summary of results from baseline of 20 °C (68 °F) fixed to fixed of 27 °C (80 °F) at full load.

27 ˚C (80 ˚F), whereas in Chicago, this same architecture
has an optimal maximum temperature of 23 ˚C (74 ˚F),
and in Miami, that temperature is only 21 ˚C (70 ˚F).
These findings may come as a surprise to many, but they
are driven by the increase in server and CRAH energy
that more than offsets the chiller savings above these
temperatures. In Miami, the economizer hours are limited
by the weather, and so the chiller savings couldn’t offset
the increase, even at 22 ˚C (72 ˚F).
BASELINE VERSUS
HIGHER FIXED TEMPERATURE
When operators think about raising temperatures
in the data center, it is commonly understood to mean
raising the temperature to a new fixed set point. Control
systems are rarely set up to handle the condition of
floating, as the analysis in this article suggests. So, the
question is: what is the impact on energy, TCO and
reliability (X-factor) if the data center temperature is to
be raised and fixed at 27 ˚C (80 ˚F)?
The server fans will always draw greater power than
the baseline scenario because the higher fixed IT inlet
temperature forces the IT fans to spin at the same faster
speed all year round. Figure 5 illustrates how the higher
fixed temperature compares to the baseline fixed temp-
erature. The findings are:
u Server energy is even higher than the floating temp-
erature scenario because the server fans are running
at the higher temperature year-round.
u Bin weather data is a significant driver in determining
whether to go to a higher operating temperature.
u Fixing at a higher temperature is always worse than
allowing the space to float to that same higher temp-
erature, because when the temperature is fixed there
are never days when the servers and CRAHs can
consume less energy.
u There is no impact on the number of economizer
hours (and therefore the chiller and dry cooler
power consumption) relative to the floating temp-
erature scenario.

May/June 2016 t 53
 FIGURE 6: Composite server
power versus inlet temperature.

CHICAGO SEATTLE MIAMI

No increase 23% savings 36% savings 13% savings

Low increase 12% savings 27% savings 1% increase

Moderate increase 12% savings 15% savings 11% increase

High increase 15% increase 10% increase 25% increase

TABLE 3: Impact on total energy (kWh) of varying CFM curves from baseline of 20 °C (68 °F) fixed to floating from 15.6 °C (60 °F)–26.7 °C (80 °F) at full load.

Table 2 on page 53 summarizes additional findings
including total energy (kWh) and pPUE. Again, this
illustrates that (1) energy is not always improved when
you raise IT temperatures, and (2) PUE as a metric alone
is insufficient.
As the inlet temperature of servers rise, the airflow
requirement and fan power increases. Figure 6 is the
composite curve from those measurements. The analysis
used this curve as the assumed ramp-up of power draw
as temperature increased.
If the server CFM requirement did not ramp up as
temperature increased (i.e., if the curve was flat), the
results of this analysis would be very different. The IT
equipment’s behavior at elevated temperatures is what
offsets the chiller energy savings, making it a complex
analysis. A flat curve would mean higher temperatures
are always better because you gain savings through
economization with no energy penalty on the CRAH
and server side.
To illustrate the impact that the CFM curve has on
the overall results, a sensitivity analysis was performed
(Table 3) holding chilled water flow constant and varying
the CFM rise as a function of temperature from flat (i.e.,
no rise) to the highest rise. The following occurs as it
moves to a steeper curve:
u Server fan power becomes a greater penalty because
power is proportional to the cube of the shaft speed.
u Number of CRAHs needed increases because more
airflow is required.
u CRAH energy increases because more airflow is
required.
u Economizer hours go down because colder chilled
water is required to make up for the decrease in
CRAH deltaT and the associated decrease in CRAH
coil effectiveness.
In all three cities, the IT equipment behavior is a key
driver to the overall energy impact of going to higher
(floating) temperatures. This illustrates the importance
of understanding the behavior of your IT gear and
analyzing the data center holistically before making
operational changes.

54 u ICT TODAY
 FIGURE 7: ASHRAE’s offset the warmer temperature-
X-factor as a function of
hours, but in a more tropical
IT inlet temperature.
environment (like in Miami),
there are not as many cool
temperature-hours to counter
those above 20 ˚C (68 ˚F).
When comparing the
baseline scenario to the higher
fixed temperature of 27 ˚C
(80 ˚F), there is a 31 percent
increase in failures. This is
regardless of location because,
now, the IT equipment is
exposed to the same higher
temperature year-round.
Another common discus-
sion point when it comes to
reliability implications of raising
IT temperatures is what happens
FIGURE 8: X-factor as a function of floating temperature for Chicago, Seattle, and Miami. in the event of a power outage.
If a data center is at a higher
initial temperature, there is
IMPACT ON RELIABILITY less ride-through time if the
The analysis thus far has focused on the optimal temperature in terms of cooling system is down before
energy and TCO savings, but reliability is another factor that must be considered things overheat and crash the
when selecting the operating temperature(s). X-factor, a metric published IT equipment.4
by ASHRAE TC9.9 committee, is the ratio of failure rate at a given dry bulb Unfortunately, today there
temperature to the failure rate at 20 ˚C (68 ˚F)—see Figure 7. seems to be a lack of quantified
The data illustrate that, relative to the failure rate of servers at 20 ˚C (68 ˚F),
data on the subject of reliability
there will be an increase in failures as the operating temperature rises. Therefore,
implication.
simply raising a fixed set point temperature will always decrease reliability if the
servers follow the curve of Figure 7.
Floating temperatures up and down is the only way to maintain reliability. For
ALTERNATIVE
example, if a data center was at 16.1 °C (61 ˚F) for half of the year (X-factor = 0.8)
SCENARIOS
The analysis presented
and 24 °C (75 ˚F) for the other half of the year (X-factor = 1.2), the average X-factor
above was based on a particular
would equal 1. In other words, there would be no impact on failures overall.
architecture with particular
Figure 8 demonstrates the impact that the maximum float temperature has
on X-factor for each of the cities analyzed. This data shows that in Chicago, assumptions. Two key varia-
floating the IT environment up to 23.3 ˚C (74 ˚F) enables cost savings without tions are addressed below
any reliability penalty; beyond this temperature, there will be an increase in because they are common
failures relative to the baseline. For Seattle, this temperature is 21.1 ˚C (70 ˚F), occurrences in today’s data
and for Miami, it is 20 ˚C (68 ˚F). This, again, is driven largely by the bin weather centers: oversized CRAHs and
data. If there are a lot of cooler temperature-hours (like in Chicago), they can lightly loaded data centers.

4
See Schneider Electric White Paper 179, Data Center Temperature Rise During a Cooling System Outage

May/June 2016 t 55
 FIGURE 9: Effect of CRAH oversizing
on 10-year energy cost floating from
15.6 °C (60 °F) to 27 °C (80 °F).

What if the CRAHs were oversized? What if the data center was
In the analysis that was described, it was assumed that the CRAH airflow only at 50 percent load?
was perfectly matched to the IT server airflow requirement, which is the best This is certainly a valid ques-
case from a capital expense perspective. However, this almost never happens tion, as most data center capacities
in practice because there is always some portion of the cool air that bypasses are based on uncertain future
the IT equipment inlets. In an actual data center, the installed CRAH airflow loads, resulting in systems that
capacity is always greater than that required by the IT equipment to ensure are under-utilized in practice. The
that all IT equipment receives the proper amount of cool air. Some of this majority of data centers operate
oversizing may be intentional, as a safety margin or for redundancy, and some between 30 and 60 percent load.
is accidental because of difficulty in forecasting loads or shrinking loads due to The same analysis described
virtualization. Uptime Institute assessments5 have found this CRAH oversizing before was run, but with the 1 MW
to be, on average, 2.6 times that required by the IT equipment. This oversizing data center loaded to 50 percent
is obviously a capital expense penalty but can actually reduce the energy (500 kW), and an additional
consumption compared to the ideal “perfectly matched” case. 25 percent CRAH capacity. The
This is due to the fan laws (sometimes referred to as the cube losses) where results are illustrated in Table 4 on
fan power is proportional to the cube of the fan shaft speed. When CRAH page 57 (baseline versus floating
airflow is oversized, the variable speed fans operate at a lower CFM (i.e., lower temperatures) and Table 5 on page
speed), therefore consuming less energy. An analysis was conducted of the 10- 57 (baseline versus fixed higher
year cooling energy implication of oversizing the CRAH airflow by 25, 50, 75, temperature).
and 100 percent while floating the IT inlet temperature up to 27 °C (80 °F). When temperatures are
Figure 9 shows that, while all three cities experienced an energy reduction as floated in a 50 percent loaded data
the oversizing increased, Miami exhibited the largest energy reduction (steeper center, the savings (as a percent
slope). This is because Miami experienced a limited number of hours at colder of the energy at the fixed baseline
temperatures where the fans could reduce their speed. Therefore, a reduction in temperature) improves. The
fan energy by oversizing the CRAH units was realized for nearly all bin hours. majority of the additional savings
Note that this CRAH oversizing comes with an increase in capital expense that comes from having less chiller
typically exceeds the 10-year cooling energy savings. While some oversizing energy which is the result of more
helps prevent hot spots in front of IT equipment, this practice must be balanced free cooling hours. This happens
with proper air management practices. because a dry cooler that is half
loaded is capable of attaining the

5
https://uptimeinstitute.com/uptime_assets/c7f39bad00527fa4e2207a5f1d5dfc1f8295a0a27287bb670ad03fafbdaa0016-00000web4.pdf

56 u ICT TODAY
 CHICAGO SEATTLE MIAMI

Total energy (kWh) 12.7% savings 10.9% savings 0.2% increase

pPUE (cooling only) Improves from 1.092 to 1.075 Improves from 1.091 to 1.079 Improves from 1.157 to 1.138

Temperature range with lowest TCO 27 °C (80 °F) 27 °C (80 °F) 21 °C (70 °F)

X-factor Improves from 1 to 0.94 Improves from 1 to 0.87 Worsens from 1 to 1.27

TABLE 4: Summary of results from baseline of 20 °C (68 °F) fixed to floating from 15.6 °C (60 °F)–27 °C (80 °F) for 50 percent loaded data center.

CHICAGO SEATTLE MIAMI

Total energy (kWh) 5.3% increase 10.6% increase 1.8% increase

pPUE (cooling only) Improves from 1.092 to 1.075 Improves from 1.091 to 1.080 Improves from 1.157 to 1.138

X-factor Worsens from 1 to 1.309 Worsens from 1 to 1.309 Worsens from 1 to 1.309

TABLE 5: Summary of results from baseline of 20 °C (68 °F) fixed to fixed of 27 °C (80 °F) for 50 percent loaded data center.

chilled water temperature earlier in the year (smaller
approach temperature).
These savings are attainable if the data center
temperature can float. In practice, this is almost never
done because control systems are not set up to adjust
temperatures dynamically/automatically. Table 5 shows
the results of the 50 percent loaded data center when
the IT space is raised to a fixed temperature of 27 ˚C
(80 ˚F). Compared to the baseline fixed temperature,
this represents an energy penalty in all three cities
analyzed. In addition, this represents an increase in the
X-factor, since the IT equipment is exposed to a constant
higher temperature. As mentioned earlier, this can be a
reliability concern. In all scenarios, PUE improved which
points to the limitation of using this as a sole metric in
making decisions.
There are several factors that go into the percent
improvements/penalties shown in these tables. A 50
percent loaded data center has an oversized dry-cooler
which enables it to get more economizer hours, which
means less time on chiller. This impacts not only the dry
cooler and chiller energy, but also (in the floating case)
the IT kWh penalty. These changes are location-specific
(BIN data-specific), and as these dynamics change, the big
drivers to the total energy also change.
For this reason, the results in this article might seem
counter-intuitive. Remember that all results shown are
relative savings/penalties compared to the baseline for
the particular location and load.
RECOMMENDATIONS
The analyses in this article demonstrate that there are
many variables that influence cost savings (or penalty),
and that raising temperatures is not always a good thing.
Before making temperature changes to a data center, it
is important to have a solid understanding of the design
conditions, system attributes, load and so on. Follow these
recommendations before raising data center temperatures:
u Air management practices such as containment and
blanking panels must be in place before attempting
to increase IT inlet temperatures. This will avoid
creating hot spots.
u Make sure the behavior of the IT equipment is
understood (power consumption and CFM require-
ment) as temperatures are raised. Ask the IT vendors
for this information.

May/June 2016 t 57
u Consider whether the BIOS settings of the IT equip-
ment can be adjusted to optimize performance at
higher temperatures. This requires a higher level of
collaboration between facilities and IT departments.
u X-factor predicts a relative increase of failure rates,
but work with IT vendor(s) to determine if the actual
rate is significant enough to be a concern.
u Since data centers are not solely made up of servers,
make sure the reliability impact on other equipment
like storage and networking is understood.
u Ensure that cooling architecture can operate at
elevated temperatures (i.e. some chillers cannot run
at higher chilled water temperatures).
u Make sure that growth plans comprehend the
potential negative energy impact of increasing
IT inlet temperatures. In other words, a savings at
50 percent load might actually be a cost penalty at
80 percent load.
u Model out how much total energy may be saved by
raising temperatures versus other optimization
strategies. Software is available to help analyze the
system dynamics of your specific data center. This is
critical because every data center will behave
differently.
u When evaluating changes, be sure to look at total
energy consumption as a metric, as PUE alone can
be misleading.
CONCLUSION
Data center operators are struggling with the decision
of raising temperatures in their IT space. Is it safe to
do this? What is the right temperature? Is it worth the
increased risk? These are some of the questions they
are faced with. This article explains the implications
of making the choice to raise IT temperatures. It is
important that the architecture be fully understood and
that a complete analysis is done before choosing the
operating points. This analysis demonstrated that:
u The cooling architecture and geographic location
(specifically the temperature profile of the climate)
has a significant impact on the optimal IT tempera-
ture set point.
u The shape of the server fan and CFM curve are
key drivers.
58 u ICT TODAY
u While raising temperatures improves the chiller
efficiency (by increasing economizer hours), those
savings can be offset by an increase in IT energy
consumption, as well as the air handlers.
u Operating conditions like percent load and CRAH
oversizing/redundancy influence whether you see a
savings or cost penalty.
u Do not assume that raising the temperature is always
a good thing. Understand the specific system
dynamics completely before making changes.
u Cooling architectures that use direct and indirect air
economizer modes will likely perform better than the
packaged chiller architecture that was analyzed in
this article. t
AUTHOR BIOGRAPHIES:
Wendy Torell is a Senior Research Analyst at Schneider Electric’s
Data Center Science Center. In this role, she researches best practices
in data center design and operation, publishes white papers and articles,
and develops TradeOff Tools to help clients optimize the availability,
efficiency, and cost of their data center environments. She also consults
with clients on availability science approaches and design practices to
help them meet their data center performance objectives. She received
her B.S. in Mechanical Engineering from Union College in Schenectady,
New York and her M.B.A. from University of Rhode Island. Wendy is an
ASQ Certified Reliability Engineer. She can be reached at wendy.torell@
schneiderelectric.com.
Kevin Brown is the Vice President of Data Center Global Solution
Offer & Strategy at Schneider Electric. Kevin holds a B.S. in Mechanical
Engineering from Cornell University. Prior to this position at Schneider
Electric, Kevin served as Director of Market Development at Airxchange,
a manufacturer of energy recovery ventilation products and components
in the HVAC industry. Before joining Airxchange, Kevin held numerous
senior management roles at Schneider Electric, including Director,
Software Development Group. He can be reached at kevin.brown@
schneiderelectric.com.
Victor Avelar is the Director and Senior Research Analyst at Schneider
Electric’s Data Center Science Center. He is responsible for data center
design and operations research, and consults with clients on risk
assessment and design practices to optimize the availability and
efficiency of their data center environments. Victor holds a B.S. in
Mechanical Engineering from Rensselaer Polytechnic Institute and
an M.B.A. from Babson College. He is a member of AFCOM. He can
be reached at victor.avelar@schneiderelectric.com.
New Brand, New Game
New beginnings with traditional values

Showing results for Hyperline Systems

Safety
Sales Support
Self-Supporting
Shielded
Steel Armoured
Structured Cabling
Systems Warranty

PROUD DIAMOND
Visit Hyperline at the BICSI Canadian Conference SPONSOR OF

& Exhibition in Niagara Falls, ON – Booth #201

May/June 2016 t 59
 “It was this big!”

DON’T BE FOOLED BY EXAGGERATED NUMBERS

The other guys like to boast about how fast or durable their splicers are. They always
like to compare their numbers to ours. It’s flattering and all, but we don’t really need
the attention. We have been pioneers in the splicing industry for over 39 years.
There’s a reason why other splicers are compared to ours.

Request a splicer demo and we’ll let you be the judge.

Fujikura 70S Splicer

• 14-second automated tube heater
www.AFLglobal.com/Truth • 7-second splice time
864.433.0333 • 6-direction drop proof