Академический Документы
Профессиональный Документы
Культура Документы
This guide provides a detailed overview of Ethernet technology. It presents common Ethernet implementations in service-
provider networks, the testing requirements to ensure reliable service, as well as installation and maintenance techniques.
Following its introduction in the early 1970s, the Ethernet protocol for data networking has been characterized by ever-increasing
popularity and adaptation. In recent years, Ethernet has become the predominant network access protocol, now used in over 95%
of all local-area networks.
With the advent of Gigabit Ethernet and 10 Gigabit Ethernet, this technology has matured and made its way from local-area
networks to metropolitan-area networks, and now wide-area networks, challenging traditional transport protocols such as
SONET/SDH and ATM.
www.exfo.com
Guide Ethernet.1-ang: Guide Ethernet.1AN 5/7/07 10:06 AM Page 1
Introduction
Guide Ethernet.1-ang: Guide Ethernet.1AN 5/7/07 10:06 AM Page 5
Customer C
half-inch-thick coaxial cable. Figure 1.2
Customer C shows a drawing used by Dr. Metcalfe to
Customer B Customer A Customer B
present Ethernet to the National Computer
Customer A Conference in June 1976.
Figure 1.2 — Ethernet VLANs provide secure virtual networks to individual customers across a MAN
IEEE 802.3 was also known as thicknet, since it specified 10 mm coaxial cable to connect devices in a bus topology. In 1985, the IEEE published
the 802.3a (thinnet) standard, which made it possible to use less expensive 5 mm coaxial cable, resulting in an immediate increase in popularity.
Since 1985, all Ethernet LAN equipment is built according to the IEEE 802.3 standard, although the vast majority of Ethernet frames sent still use
the DIX Ethernet 2 format. The differences between Ethernet 2 and IEEE 802.3 are subtle and are transparently managed by network equipment;
the frame type is not a concern during practical use or testing.
Due to economies of scale, Ethernet equipment has historically been more cost-effective than competing technologies. Manufacturers are now
offering economic solutions for Ethernet switching and routing, virtual LAN (VLAN) services, MPLS, and IP forwarding, as well as traffic-
management features that allow service providers to fulfill various service agreements. Ethernet services require less time for commissioning and
provisioning than legacy services, and can even be provisioned remotely, resulting in additional savings for service providers. Since almost all LANs
use Ethernet, they can easily be integrated with Ethernet-based carrier networks since Ethernet equipment performs bit-rate conversion and
statistical multiplexing.
Ethernet is flexible in that it can be used with many different infrastructures and protocols. Ethernet services can be offered over existing ATM
architectures. In next-generation SONET/SDH networks, Ethernet frames can be encapsulated into Generic Framing Procedure (GFP) frames and
transported through SONET/SDH channels. Ethernet can also be transmitted in its native format over dark fiber or on a DWDM wavelength,
or using free-space optics (FSO).
Other advantages of Ethernet include dynamic bandwidth commissioning (ability to increment bandwidth to subscribers on an as-needed basis)
and a wide range of services based on Ethernet standards for wide-area networking (virtual private LAN, E-Line services, etc.). Ethernet services
will be described in detail in Chapter 3.
Frame Check
10101... Sequence
CRC-32
Ethernet
Known as MAC addresses (short for media access control), these are
S Destination Source hardware addresses that uniquely identify each node on a network.
Preamble O Type DATA FCS
F Address Address
7 1 6 6 2 46-1500 4
Each address is associated with a user (host) device, more specifically with an interface on that device. It consists of a 6-byte (48-bit) field, which
must contain a globally unique number.
The address is divided into two sections—the first (sometimes known as the organizationally unique identifier or OUI) is allocated to network
interface card (NIC) manufacturers by the IEEE; the other section is like a serial number allocated to each NIC by the manufacturer himself. Shown
below is a typical MAC address in its various formats.
0000 1000 0000 0000 0000 1100 1001 1110 0001 0111 1000 1010 Binary
0 8 0 0 0 c 9 e 1 7 8 a Hex
08 — 00 — 0c — 9e — 17 — 8a Sometimes
Written
08 : 00 : 0c : 9e : 17 : 8a
08 . 00 . 0c . 9e . 17 . 8a
0800 . 00c 9e . 178a
IEEE Assigned Manufacturer Assigned
Figure 2.4 — Sample MAC address in three common formats
The IEEE manufacturer number allocation forms the most significant part of the MAC address; and this, combined with the serial number, provides
a globally unique address for every manufactured NIC.
Various formats exist to represent MAC addresses, but they all break down into a binary number. Although the IEEE OUI is assumed to be 24 bits
long, the two most significant bits have another purpose:
bit 47 — when set to 0, indicates an individual address; when set to 1, indicates a group (multicast) address.
bit 46 — when set to 0, indicates the universally administered global addressing IEEE scheme; when set to 1, indicates a locally
administered addressing scheme.
A MAC address of FF-FF-FF-FF-FF-FF (all ones) is reserved for use as a broadcast address.
Unfortunately, the MAC system is based on what is known as a flat addressing scheme, as
Destination Source
Address Address opposed to a hierarchical scheme. For example, a flat scheme would be the equivalent of a street
6 6
full of houses with names instead of numbered addresses; the only way of locating a specific house
would be to inspect them all individually. With a hierarchical scheme, on the other hand, the
sequential numbered addresses make it much easier and quicker to locate the position of a specific
00-80-0c-9e-17-8a destination. This is why MAC addressing is only used in confined (local) areas.
A B Here is the address shown above in context of the destination address. Both Ethernet and 802.3
A: Address specified by IEEE
use a similar address format.
B: Address specified by manufacturer
Ethernet TYPE field — identifies the protocol type of the data field for
IP
Network Layer forwarding to the appropriate higher-layer protocol.
08-00 06 802.3 LENGTH field — indicates the length of data (minus padding) in the
Type Snap
data field.
Ethernet 802.3/802.2 Data-Link Layer
2.2.5 802.3 Logical Link Control Header (LLC)
The IEEE, having written their 802.3 specification according to a LENGTH
00-03-01-xx-xx-xx
Or FF:FF:FF:FF:FF:FF field, were faced with finding another way to represent the function of the
Coax, UTP/STP, Fiber Optics Physical Layer
TYPE field. They did this by creating another protocol (LLC) that ‘sits’ directly
above the MAC layer (but still within the Layer 2 data-link layer).
Figure 2.7 — Type and snap fields defining IP Layer 3 payload
At the LLC level, Ethernet functionality is not evident, so this protocol, defined
in the 802.2 specification, can function on other LAN types as well (e.g., token
ring, FDDI, etc.), forming a common layer between them.
LLC (802.2)
This 802.2 specification is known as the logical link control (LLC) layer.
DSAP SSAP Control
1 1 1-2
IEEE 802.3
Preamble Destination Source Length 802.2
S
O
Address Address
F Header DATA FCS
7 1 6 6 2 46-1500 4
Devices detecting a collision will generate a predetermined jamming signal to tell all other devices that a collision has occurred and will then
withdraw for a while (referred to as a Backoff Strategy, in which the specific withdrawal time period is governed by a randomly generated number),
before going back to the Carrier Sense phase and trying again.
2.13 CSMA/CD Transmission Flow
Data Station
Ready to Send 2.13.1 Full-Duplex Ethernet
New Attempt Wait According to
Backoff Strategy In full-duplex Ethernet, devices at each end of a link can send and receive data simultaneously.
Sense One advantage of this approach is that the full-duplex link can theoretically provide twice the bandwidth
Channel Channel Busy
of normal (half-duplex) Ethernet. The full-duplex mode of operation requires that each end of the link
Channel Free
connect only to a single device, such as a workstation or a switched hub port.
Transmit Data Transmit A device at the end of a full-duplex Ethernet link does not have to listen for other transmissions or for
Sense Channel Collision Jam Signal
Detected collisions when sending data. Therefore, there is no need to adhere to the original Ethernet medium
No collision Detected
Transmission
access control system (CSMA/CD);
Complete
The 10Base-T, 100Base-TX, and 100Base-FX and 1000Base-T/X signaling systems support full-duplex
Figure 2.16 — CSMA/CD algorythm operation; they have transmission and reception signal paths that can be simultaneously active.
Flow control is defined for neighboring devices on a point-to-point link; i.e., flow control is not implemented end to end. If a receiving station
becomes congested, it sends a pause frame to the transmitting station specifying the amount of time for which the transmitting station should stop
sending packets. The transmitting station waits for the specified time before sending additional packets. This effectively reduces throughput
between the two devices and results in much less retransmission, as the packet loss that occurs from lack of flow control is no longer an issue.
A second form of flow control, called XON/XOFF, toggles traffic flow by sending one pause frame to stop traffic flow and a second to request a
resumption of traffic (typically when the receiving buffer has emptied).
2.13.3 Pause Frame Format
S
PreambleO Destination Source
Preamble 8808 DATA FCS The pause frame is a standard Ethernet frame that is sent where the type field
F Address Address
is specified as 8808. The pause time is specified in the data payload as
0100 0-0xFFFF PAD (0x0000)
shown below.
802.3x Pause 2.13.4 Auto-Negotiation
OpCode Time
There is an optional part of Ethernet allowing two devices to negotiate the
Figure 2.17 — Pause frame format and location best possible connection between them. Auto-negotiation devices exchange
information about their range of link speeds, possible modes of operation
(full-duplex or half-duplex), and whether they support flow control. For 1000Base-T links, auto-negotiation also includes master clock support.
Flow control can be implemented symmetrically or asymmetrically (if one device has manually configured and/or fixed transmission settings).
On copper connections, the auto-negotiation process takes place using a modified version of the normal link pulse (NLP) signals used to verify
link integrity, called the fast link pulse (FLP) signals. It should be noted that NLPs and FLPs are specified only for twisted-pair media using
eight-pin connectors, such as 100Base-TX over unshielded twisted-pair wire.
Optical links use a special order-set control symbol instead of FLPs to carry auto-negotiation information.
Devices on a network that are capable of auto-negotiation find their highest-performance mode of operation based on a set of negotiation rules,
shown in the table below.
Auto-Negotiation Priority Rules
A 1000Base-T
B 100Base-TX full-duplex
C 100Base-T4
D 100Base-TX
E 10Base-T full-duplex
F 10Base-T
The negotiation process works its way from F to A until the highest value match is achieved. Switches and hosts can override auto-negotiation
with manual settings. 10 GigE does not support auto-negotiation.
2.13.5 Link Aggregation
Link aggregation (or trunking) is another Ethernet feature only applicable to the full-duplex operation. It provides increased link availability and
bandwidth between two Ethernet stations by allowing multiple physical links to be combined as a single "logical" link. The link-aggregation
specification is specified in 802.3ad.
Prior to link aggregation, it was difficult if not impossible to have multiple links between two Ethernet stations; the spanning tree protocol (STP)
algorithm used in Ethernet bridging (802.1D) disables parallel paths to prevent loops in the network.
Link aggregation allows multiple links between any two Ethernet stations if they consist of two switches, a switch and a server, or a switch and an
end-user station. The following advantages are provided:
• Bandwidth can be increased by combining available lines (for example, two 10 Mb/s links can result in a 20 Mb/s link)
• Load balancing can be performed by distributing traffic across the multiple links. One line can be dedicated to high-priority traffic, if required.
• Redundancy is provided by the multiple links
Link aggregation operates by adding a new layer of function between the Ethernet MACs and the higher-layer protocols above.
The link-aggregation function is completely transparent to all higher-layer protocols and functions, including the spanning tree algorithm, VLANs,
SNMP and routers.
Aggregation can only be performed if:
• Links are point-to-point (no multipoint permitted)
• Links must operate in full-duplex mode
• Links being aggregated run at the same data rate (e.g. 10 Mb/s, 100 Mb/s, or 1 Gb/s)
WAN PHY runs at a data rate compatible with SONET OC-192c/SDH STM-64c (9.5846 Gb/s). The WAN PHY is designed to support
connections with SONET/SDH circuit-switched networks. It adds a WAN interface sublayer (WIS) to the LAN PHY. The WIS takes the
data payload and encapsulates it with its embedded SONET/SDH frame into a simplified SONET OC-192c/SDH STM-64c
(concatenated) frame. Because of the SONET/SDH overhead and the 64B/66B encoding, the actual data rate supported is lower than
the LAN PHY, so the WIS has a mechanism to pad and buffer data to the 10 Gb/s MAC data rate.
WAN PHY is designed to bridge asynchronous data media and synchronous TDM transport networks, which allows 10 GigE to be
transparently carried across today’s TDM infrastructure. It is 100% compatible with current DWDM networks carrying OC-192/STM-64
and 100% compatible with OTN (DigiWrapper) networks carrying OC-192/STM-64.
However, it should be understood that this WAN interface is not true SONET. The "S" in SONET (or SDH) stands for synchronous; that
is, all points on the network are synchronized to an accurate central master clock. Ethernet is an asynchronous system in which each
receiving device derives clock and data from the incoming stream and re-times the outgoing characters with a local clock. The 10 Gigabit
Ethernet output from a device with a WAN PHY does not connect directly to a SONET/SDH ring as it requires an access device.
WAN PHY has some SONET/SDH features, but does not support the full SONET/SDH standard:
• Supports only the SONET/SDH overhead features required for fault isolation
• Ignores line and section DCC
• Ignores local and express orderwire
• Supports pointer processing to allow carriage over future OC-768/STM-256 backbones
Telco-like features supported on 10 GigE WAN PHY:
• WAN PHY has facility-loopback capabilities
• WAN PHY has threshold for severely errored seconds
• WAN PHY has embedded test pattern generation/detection capabilities (PRBS-31)
Location A
Data Routing
Center Switch
10 GE
WAN PHY Transponder
Routing Location B
Switch 10 GE
WAN PHY
Transponder DWDM
Optical Network
Transponder 10 GE
WAN PHY
Routing
Switch
Remote
Servers
Location C
OC-766 Mux
Location A OC-766 Mux
National
Backbone
10 GE
WAN PHY
Routing
Switch 10 GE OC-766 Mux
WAN PHY
DWDM
Optical Network
Figure 2.19 — 10 GigE WAN PHY applications: OTN and OC-768 compatibility
Edge Routers
OC-48 10 GigE WAN link
Ethernet Applications
Guide Ethernet.1-ang: Guide Ethernet.1AN 5/7/07 10:06 AM Page 36
3. Ethernet Applications
Ethernet is widely used in local-area networks (LANs) as well as in different types of carrier networks — access networks, metropolitan-area
networks (MANs) and long-haul or wide-area networks (WANs).
Hub A Repeater joins two network segments together to overcome cable-length limitations. The role
of this element is to simply regenerate signals. A maximum of four repeaters are allowed between
Host any two terminating clients (computers).
A Hub is a multiport repeater that joins devices to a common backplane. Within the hub, the data
goes through a repeater process.
Workstation
Host Host
Hubs can have different
Hub Hub
interfaces for different cable
types; they can offer rate
Shared backplane in each hub conversion and, if software is
Hub Uplink
added to them, they can become
Workstation Host managed devices. If hubs are
connected to each other to
Hub Hub increase the number of available
Host ports, they must also follow the
four-consecutive-repeater limit.
Workstation Workstation
Figure 3.3 — LAN using multiple hubs to add users to a common network
All devices connected to a hub (or via interconnected hubs) are capable of colliding with each other so they are in the same collision domain
(CD), allowing only one device to transmit at a time. CSMA/CD is a very efficient media access scheme under light to moderate traffic loads.
As the traffic volume gets heavier, however, a significant amount of network transmission time becomes consumed by collisions, and networks
need to be segmented (using a switch) to restore traffic efficiency.
Host The use of hubs is declining because switches are now just as inexpensive and
High Speed, High they offer a much more efficient way of constructing the LAN.
Bandwidth Backplane
Switches and Bridges
A bridge divides a network into two segments. It restricts traffic to a
single segment unless the device needs to access a device on the other
segment (known as bridging). A hub cannot segment a network since
Switch segmenting is performed using MAC addresses, which are not
Host
understood by hubs; they are simply Layer 1 (repeater) devices.
A strategically placed bridge can significantly reduce collisions by
segmenting workgroup traffic flow.
Host
A switch is a multisegment bridge. Each segment is connected to a port,
Workstation which can be bridged to any other segment (port). Switches have a large
bandwidth backplane, allowing devices on two segments to be
temporarily connected at what is called wire-speed, while simultaneously
Figure 3.4 — A typical office LAN topology using a Layer 2 switch enabling similar bridging between other device pairs.
If every client on a network is interconnected through a switch, collisions
do not occur and maximum possible throughput is enjoyed by all clients.
Switches allow multiple hosts to transmit at the same time, without
contention for bandwidth. Switches can be VLAN tag-aware, permitting
38 Ethernet Reference Guide EXFO www.exfo.com
Guide Ethernet.1-ang: Guide Ethernet.1AN 5/7/07 10:06 AM Page 39
Metropolitan Networks
Long-Haul/ A Router is a multilayer switch used to direct Layer 3 (commonly Internet
metro
(MANs) Protocol, or IP) traffic. By examining the IP address, a switch determines
Intra-metro Intra-metro
connection Inter-metro connection the corresponding MAC address so that the Layer 3 traffic can be
connecton
correctly forwarded through a Layer 2 Ethernet network. Address
IP Router Frame
resolution protocol (ARP) is the standard mechanism used to correlate
SONET Realy
ATM
ESCON
Fiber
MAC and IP addresses in routers.
Ethernet Gigabit
Switch LAN Channel Ethernet
Residential x-DSL or
cable modem networks
PSTN/cellular Regional ISP Corporate
enterprise clients
3.2 Ethernet in Access Networks
Personal computers and LANs are usually connected to the outside world via
Figure 3.5 — Interconnection of access networks, MANs and WANs
an access network (see figure below). An access network, commonly referred to as the first mile or the last mile depending on the context, covers
a relatively small geographic area and connects subscribers, such as corporations, government offices, educational institutions and residential
customers, to a metropolitan-area network (MAN). MANs in close proximity may be directly interconnected. Long-distance connections between
MANs are provided by long-haul telecom networks to create wide-area networks (WANs).
Access networks use a variety of media such as twisted pairs of copper wire, hybrid fiber coax (HFC) or, optical fiber in FTTx and passive optical
networks (PONs). Access rates range from sub-rate channels such as DS1, DS3, OC-3, OC-12, Fast Ethernet and Gigabit Ethernet,
to full-wavelength capacities such as OC-48 and OC-192.
Access networks use a variety of protocols including ATM, frame relay, SONET/SDH, Ethernet and MPLS. Ethernet services can be delivered over
any of these access technologies, but the most cost-effective and simple method is to deploy Ethernet services directly over Ethernet access lines.
Ethernet access networks typically run at 1000
Customer C
Mb/s (Gigabit Ethernet, or GigE). Traffic is
Customer A Customer B
Customer B Customer A delivered to the customer from the tributary side
Store Customer B of an access router or Layer 2 switch. Rate limiting
Customer C is commonly used to limit the bandwidth to the
Store
Figure 3.6 — VLANs provide secure virtual networks to individual customers across a MAN
IP Video Figure 3.7 and Figure 3.8 illustrate the two basic
service types and how they are used to
interconnect geographically separated LANs by
Point-to-Point
Ethernet Virtual Connections interfacing customer edge (CE) equipment to
(EVCs) the MAN. Note that each customer requires only
Servers
IP Voice one CE, regardless of the number of Ethernet
virtual connections (EVCs) involved. It would even
Data be possible to make a multipoint-to-multipoint
CE
IP PBX
connection and several point-to-point connections to
MAN IP Video the same CE. In all cases, the Ethernet service ensures
CE that frames are delivered to the correct destination.
IP Voice
Data
CE
3.6.5 ATM
Ethernet can be carried over existing ATM networks by using the information in each Ethernet frame to map the frame to an appropriate ATM virtual
circuit and service class. Ethernet services can thus be provided over ATM with the same QoS and resiliency as ATM. This method allows
considerable flexibility in service topology, including point-to-point and multipoint-to-multipoint services and optimal levels of oversubscription.
Greenfield deployments, however, are more likely to use IP/MPLS, rather than ATM, due to advantages in cost and scalability.
3.6.6 IP/MPLS
The first Ethernet-based carrier services were less expensive and more flexible than leased lines or frame relay services, and they also offered full
support for VLANs. However, they were not carrier class and lacked the reliability, scalability and security of traditional TDM and ATM services.
In addition, the IP routers used to provide Ethernet IP services could not perform longest prefix match lookups at wire speed, resulting in lower
performance.
Multiprotocol label switching (MPLS) is a control-plane packet-forwarding technology that is rapidly being adopted to overcome these drawbacks
and make Ethernet “carrier class” – allowing Ethernet to transmit voice and other delay-sensitive applications. MPLS also makes the entire network
easier to provision and engineer. As the name implies, MPLS technologies are applicable to multiple network layer protocol including IP, ATM and
frame relay.
MPLS helps move traffic faster by building virtual circuits, or tunnels, called label-switched paths (LSPs) across the network (see Figure 3.9) and
by using simple labels to make data-forwarding decisions. The ingress-edge label-switching router (LSR) analyzes the Layer 3 header of each
packet to determine the destination address and assigns a label to the packet. The LSRs in the core only inspect the label in order to forward the
packet to the next hop in the LSP. At each hop, the LSR strips off the existing label and applies a new label, which tells the next hop how to forward
the packet. The LSR at the egress edge removes the label and forwards the packet normally.
Label-switched paths (LSPs) can be established for virtual private networks to guarantee a certain level of performance or to route around network
congestion. LSPs are similar to circuit-switched paths in ATM or frame relay networks, except that they are not dependent on a particular
Layer 2 technology.
MPLS also provides other benefits, such as virtual private networking (VPN) to create IP tunnels through the network; traffic engineering (TE) to
provide traffic prioritization for different classes of traffic; quality of service (QoS); and bandwidth management. Another feature is fast rerouting,
which allows traffic to be quickly transferred to a pre-established backup tunnel. This is similar to SONET’s protection switching.
Frame Relay
Cirtual
ATM Circuits
IP/MPLS
MPLS-Based Universal
Port
Core
Packetized
MAN
Ethernet OC-n
OC-n OC-n
OC-n Port
Optical
SONET
When these measurements are performed, they provide a baseline for service providers to define SLAs with their customers. They enable service
providers to validate the quality of the service delivered and can provide them with a tool to create value-added services that can be measured
and demonstrated to customers. For example, these tests provide performance statistics and commissioning verification for virtual LAN (VLAN),
virtual private networks (VPNs) and transparent LAN services (TLS), all of which use Ethernet as an access technology.
The SLA criteria defined in RFC 2544 can be precisely measured using specialized test instruments. The performance verification is usually done
once the installation is complete. The measurements are taken when the network is out of service to make sure that all parameters are controlled.
4.2.1 Test Configurations
Different configurations are possible for performing the tests. These are explained below.
Dual-test-set configuration
Two test sets are required for local/remote testing, also known as head-to-head testing (see Figure 4.1). This test configuration, also known
as ”master-slave”, is ideal for:
• End-to-end testing
• Going through a routed network
This configuration, the user operates one test set, which controls
the other by designating one as the local test set and the other as the
Network remote test set. This makes it easy to determine in which direction the
under test traffic is flowing. The remote test set is controlled via the connection
under test. During the test, results from both directions are visible on
the local test set.
Figure 4.1 — Dual-test-set configuration (arrows show direction of traffic)
Dual-port configuration
In some cases, testing can be performed using one test set with two ports (see Figure 4.2). This is ideal for:
• Testing a switch locally
• Lab environment
• End-to-end testing with high port density
Port 1
Single-port loopback configuration
Customer
The single-port loopback configuration is similar to the dual-port configuration in
Network that it requires only one test set. The difference is that the traffic does not go from
Port 2 one port to another but simply loops from the transmit connection of one port to
the receive connection of the same port (see Figure 4.3). It is ideal for:
• DWDM systems
Port 1
• End-to-end testing with end devices that can loop back to the same port,
10/100 Mb/s
Layer 2 Switch either with a cable or a software loopback
Port 2 Note: Many systems (mainly Ethernet switches or routers) will not allow a
1000 Mb/s simple loopback to the same port. This is because switches forward frames
according to a destination MAC address. Routers work the same way, but use
Port 1 a destination IP address. If such a device is looped back to the same port, it will
not know where to forward the frames and will drop them. A special loopback
Customer
Network device that can correctly address the originating test set is required for these
Port 2 configurations.
Switch A Switch B
Throughput
Remote Testing
Throughput is the maximum rate at which none of
the offered frames are dropped by the device
Internet under test (DUT) or network under test (NUT).
For example, the throughput test can be used to
measure the rate-limiting capability of a switch.
The throughput is essentially equivalent to the
bandwidth and can be measured bidirectionally
using a dual-port loopback configuration or
Gigabit Ethernet unidirectionally using two test sets (see Figure 4.5).
Metro Network
Gigabit Ethernet The throughput test allows vendors to report a
single value, which has proven to be useful in the
10/100Base-T 10/100Base-T marketplace. Since even the loss of one frame in
a data stream can cause significant delays while
waiting for the higher-level protocols to time out,
it is useful to know the actual maximum data rate
that the device can support. Measurements
Figure 4.4 — End-to-end, end-to-core, and remote testing should be taken over an assortment of frame
sizes. Separate measurements should be
performed for routed and bridged data in those
devices that can support both. If there is a
checksum in the received frame, full checksum
processing should be done.
# of Bytes
Burst test procedure:
1. Send a burst of frames with minimum inter-frame gaps to the DUT/NUT
and count the number of frames forwarded by the DUT/NUT.
X STM-1
3
All test procedures are from RFC 2544 (http://www.faqs.org/rfcs/rfc2544.html).
Latency
For store-and-forward devices, latency is the time interval between frames (input and output); it starts when the last bit of the input frame
reaches the input port, and ends when the first bit of the output frame is seen at the output port. Round-trip latency is the time it takes a frame
to come back to its starting point. Variability of latency can be a problem. With protocols like VoIP, a variable or long latency can cause degradation
in voice quality.
4.4.1 GigE and 10 GigE BERT over Dark Fiber and PONs
Carriers are increasingly using dark fiber and passive optical networks
(PONs) to provide point-to-point and point-to-multipoint access links,
GigE respectively, using physical-layer fiber transceivers. When Ethernet
POP services are provided over these links, short light budgets and partial
5-10 Km
fiber faults will reduce GigE throughput. BERT over Ethernet testing
allows providers to verify the ability to fully load the GigE bandwidth
without any bit errors (see Figure 4.12).
CPE
OC-3
Figure 4.12 — Testing a GigE fiber point-to-point access link OC-12
Gigabit Ethernet
ESCON
OPTera
4.4.2 GigE and 10 GigE BERT over a DWDM Network DWDM Ring
Extension Rings
When GigE is transported transparently across DWDM networks, wavelength crosstalk and
transponder fade-out reduce GigE throughput. Ethernet BERT is required to validate error-
free GigE transmission across the DWDM ring (see Figure 4.13).
DWDM Ring
Loop
CLEC Office
Video streaming Video payload size (bytes) Video bandwidth (f/s) Ethernet frame size (bytes)
Business Quality Videoconferencing 915 107 973
NetMeeting Video LAN (videoconf) 779 77 837
NetMeeting Video DSL (videoconf) 363 65 421
NetMeeting Video 28K (videoconf) 288 5 346
PictureTel StarCast (stream) 1343 155 1401
RealAudio Radio (stream) 681 30 739
Media Player 80K (stream) 697 15 755
Media Player 20K (stream) 476 7 533
Real Video 28K (stream) 384 8 442
Table 5.1 — Common packet characteristics for multimedia applications including video
When commissioning VoIP or video services, the test set used should be able to configure streams with these parameters.
When stressing a network with multiple streams, detailed measurements are performed on one stream while controlling the priority, bandwidth and
characteristics of the others. This technique enables realistic network traffic simulation over a wide variety of typical boundary conditions.
Loopback To ensure that unidirectional effects are accounted for, it is important to use a
testing test setup that permits parameter measurement in each direction
independently. Most Ethernet testers can only perform measurements in
Upload direction
loopback configuration (see Figure 5.1). This results in upload and download
Download direction test scores being combined into a single, averaged measurement, making it
impossible to separate upload and download issues.
Remote (slave) Local (master)
Unidirectional Ethernet network It should be noted that one parameter — latency — cannot be measured
testing unidirectionally without sophisticated clock synchronization schemes
(involving GPS signals, for example). Latency is typically tested with a loopback
Upload direction configuration. All other performance metrics including throughput, burst,
Download direction frame loss, and packet jitter should be measured unidirectionally to obtain
accurate test results.
Remote (slave) Local (master)
Statistics and alarms provide a high-level view of the network in order to detect problems. By using probes or protocol analyzers on the problem segment,
the service provider will usually be able to identify, troubleshoot and resolve most problems before a customer experiences degraded performance.
By gathering traffic statistics in a database, trending patterns can be provided to network management staff. With this data, the service provider
can architect new network installations or proactively groom traffic through optimal routes to provide bandwidth to trouble areas. This approach
generates additional revenues through customer retention and a reduction in SLA performance-penalty payouts.
Test solutions based on the 802.3ah guidelines address all of the key challenges
Application
Presentation
commonly encountered with Ethernet service deployment: reliability, performance, QoS
Session LLC — Logical Link Control assurance and network troubleshooting.
Transport OAM — 802.3ah Standard for Ethernet
Key to the standard are specifications for remote loopback testing to customer
Network MPMC — Multipoint MAC Control
Data Link MAC — Media Access Control
premises. This single-ended testing technique permits proactive network management
Physical and speeds up restoration of failed services. In addition to loopback testing, 802.3ah
also provides guidelines for remote failure and event notification, link-monitoring
Figure 6.3 — Open Systems Interconnection (OSI) Reference Model, statistics, diagnostics, as well as the ability to collect historical performance
showing 802.3ah OAM sub-layer in data-link layer
parameters from 802.3ah-compliant elements. Because this standard emulates
common DS1/3 access maintenance practices, 802.3ah management solutions are
easily integrated into existing operations. As the 802.3ah standard is increasingly
adopted by system, switch and broadband equipment suppliers, remote testing and
monitoring deployment is becoming straightforward and cost-effective.
6.2.1 Access Line Management Using the 802.3ah Standard
Remote testing with 802.3ah involves three key elements:
• A demarcation device located between the provider and customer networks
Figure 6.4 — An Ethernet NIU made by ADC (customer and provider side) • A remote test-head capable of advanced performance evaluation and
commissioning functions
• Management software capable of identifying and cataloging demarcation devices
and their location to simplify test access
Customer
Remote testing can also be used for commissioning, monitoring and SLA
Premises 2 management. Remote bandwidth provisioning is a typical example of the
Ethernet Tester advantages of an 802.3ah-based Ethernet access network. When a
customer requests additional bandwidth on their access link, the related NIU
Ethernet Metro Network is remotely accessed by the provider and a message is sent to immediately
Switch SONET/SDH/DWDM increase the throughput available to the customer. The NIU is then put into
loopback mode and remote Ethernet performance tests are conducted to
NIU ensure that the expected bandwidth is available. A series of tests are then
Customer
Premises 1 performed to ensure that revised SLA parameters will be met under various
(under test) network conditions. With the new bandwidth assured and commissioned, the
Figure 6.8 — In a virtual private LAN a portable Ethernet tester can be used to test from NIU is returned to pass-through mode and the customer’s traffic passes over
one customer location to a second site where an NIU is installed the upgraded link.
www.exfo.com Ethernet Reference Guide EXFO 75
Guide Ethernet.1-ang: Guide Ethernet.1AN 5/7/07 10:06 AM Page 76
NIU 1 NIU 3
Customer Customer
A B
Glossary
Guide Ethernet.1-ang: Guide Ethernet.1AN 5/7/07 10:06 AM Page 78
7. Glossary
A B
Adapter: Also known as a network interface card (NIC). An adapter is Backbone: Part of the network that joins several local-area networks,
a circuit board installed in a computing device to connect it to a either inside a building or across a city or country. This is achieved
network. The adapter or NIC performs the hardware functions that are through a cable connection between telecommunication or wiring
required to provide a computing device with physical communication closets, floor distribution terminals or entrance facilities. In star networks,
capabilities. the backbone cable interconnects hubs and similar devices, as opposed
to cables running between hub and station. The backbone is the part of
Address Resolution Protocol (ARP): A network-layer protocol used in the communications network that carries the heaviest traffic.
TCP/IP transmission. ARP is used by end stations to determine the
physical address of other stations on the same LAN. Backoff Delay: In Ethernet transmission, the backoff delay is the length
of time that a station waits before retransmitting a frame, after a data
Asynchronous: Said of transmission in which sending and receiving collision is detected. This operation applies to carrier-sense multiple-
devices are not synchronized. Data division is indicated by data itself, access networks with collision detect (CSMA/CD; see separate entry).
which carries these signals.
Baseband: A transmission method in which one single digital signal
Asynchronous Transfer Mode (ATM): A data networking protocol uses an entire bandwidth. The unmodulated signal is sent directly over
used for high-bandwidth, low-delay, connection-oriented, packet-like the transmission medium. Baseband is simpler, cheaper and less
switching and multiplexing. sophisticated than broadband. All Ethernet media types are baseband
Auto-Negotiation: Algorithm allowing two devices (at each end of a link) except for 10Broad36, which is broadband.
to negotiate common data service functions (i.e., transmission rate, half vs.
full duplex, etc.)
Bandwidth: The range of frequencies required for proper transmission Broadband: A transmission medium whose bandwidth capacity is
of a signal. Bandwidth represents the amount of data that can be sufficient to carry multiple voice, video or data channels simultaneously.
transmitted through a communications channel in a fixed amount of Each channel is modulated to a different frequency bandwidth and
time. For digital devices, it is usually expressed in bits (or bytes) per occupies a different place on the transmission medium; the signals are
second, whereas for analog devices, it is expressed in cycles per then demodulated to their original frequency at the receiving end.
second, or in hertz (Hz). The greater the bandwidth, the greater the 10Broad36 is the only broadband Ethernet media type. All other
information-carrying capacity and the faster the speed. A continuous Ethernet media types are considered baseband.
frequency range starting from zero is said to be baseband, while a
Broadcast: The act of sending a frame to all network stations. Also
range starting substantially above zero is considered broadband.
describes the class of media (designed especially for CSMA/CD
Bit: One binary digit Ethernet) in which all stations are capable of receiving a signal
transmitted by any other station.
Bit Error Rate (BER): A measure of data integrity referring to the
number of digital highs that are interpreted as lows (and vice versa), Broadcast Address: A multicast address that identifies all the stations
divided by the total number of bits received. The BER ratio is often on a network.
expressed as a negative power of ten.
Broadcast Domain: A restricted area that allows all connected
Bridge: Specified in IEEE 802.1D standard, a bridge is a device that devices to transmit and receive information from each other. The
connects two or more networks at the data-link layer (Layer 2). Bridges devices are interconnected through bridges, allowing them to share the
are not part of the collision domain; i.e., they may be used to split a transmission medium and, consequently, the data.
network into multiple collision domains.
Byte: A group of 8 bits. Also known as an octet.
Cyclic Redundancy Check (CRC): Error-checking technique used Destination MAC Address: Address identifying the station or stations
to ensure the accuracy digital-code transmission over a on a LAN to which a frame is being sent.
communications channel. The transmitted signals are divided into
predetermined lengths which, used as dividends, are divided by a DIX: Acronym identifying the three companies that released the original
fixed divisor. The remainder of the calculation is appended onto and Ethernet specification in 1980: Digital, Intel, and Xerox.
sent with the message. At the receiving end, the computer
recalculates the remainder. If it does not match the transmitted Duplex: Circuit used to transmit signals simultaneously in both directions.
remainder, an error is detected.
D E
Data-Link Layer: Layer 2 of the OSI reference model. This layer takes Electronic Industry Association (EIA): An association of
data from the network layer and passes it on to the physical layer (layer manufacturers and users that establishes standards and publishes test
1). The data-link layer is responsible for transmission and reception of methodologies. Formerly known as RMA or RETMA.
Ethernet frames, 48-bit addressing, etc. It includes both the media
access control (MAC) and logical link control (LLC) layers. Ethernet Version 2: The original Ethernet specification produced by
Digital, Intel, and Xerox (DIX) that served as the basis for the IEEE
Dense Wavelength-Division Multiplexing (DWDM): A technology 802.3 Ethernet standard.
that enables a single optical fiber to carry multiple data channels (or
wavelengths). Commercial DWDM systems can have as many as 160 Excessive Collision Error: Error that causes frame loss. This type of
separate channels. error occurs when a station receives 16 consecutive collisions while
attempting to transmit a single frame; then the frame is dropped due to
the excessive collisions.
G H
Gb/s or Gbps: Gigabits per second. One Gb/s equals one billion bits Half-Duplex: Data transmission over a circuit capable of transmitting
per second. in either direction, but not simultaneously. For Ethernet, the CSMA/CD
method is a half-duplex protocol.
Generic Framing Procedure (GFP): Traffic adaptation protocol
providing convergence between packet-switched and transmission Headend: The equipment in a cable system which receives the various
networks. GFP elegantly maps packet-based protocols such as program source signals, processes them, and retransmits them to
Ethernet, Fibre Channel, FICON, ESCON, and various forms of digital subscribers.
video into SONET/SDH, typically using virtual concatenation to
provide right-sized pipes for data services. Hub: A device at the center of a star topology network. Hubs can be
active (where they repeat signals set to them) or passive (where they
Giants: Giants are frames that are longer than the maximum Ethernet do not repeat but merely split signals sent through them). Hub may
size (giant frames > 1518/1522 bytes with bad FCS, whereas oversize refer to a repeater, bridge, switch, router, or any combination of these.
frames > 1518/1522 bytes with good FCS). Giant packets usually
occur when you have a jabbering node on your network; i.e., a node
that is continuously transmitting, or transmitting improperly for short I
bursts – probably due to a bad transmitter on the NIC. Giants can also
IEEE: Institute of Electrical and Electronics Engineers. A professional
be caused by packets being corrupted as they are transmitted, either
organization and standards body. The IEEE Project 802 is the group
by the addition of garbage signals, or by the corruption of the bits that
within IEEE responsible for LAN technology standards.
indicate frame size.
IEEE 802.1: The IEEE standards committee defining high-level
Gigabit Ethernet (GigE): A version of Ethernet that operates at 1 Gb/s
interfaces, network management, internetworking, and other issues
(1000 Mb/s).
common across LAN technologies.
Group Address: An address specifying a group of logically related
stations on a network. Also called a multicast address.
Link Segment: In Ethernet, a point-to-point segment that connects two Mb/s or Mbps: Megabits per second. One Mb/s equals one million
and only two transceivers at its endpoints. bits per second.
Local-Area Network (LAN): A term used to refer to a form of Media: Wire, cable, or conductors used for transmission of signals.
networking technology that implements a high-speed, relatively short
distance form of computer communications. Ethernet is one type of LAN. Medium Access Control (MAC): A mechanism operating at the data
link layer of local-area networks which manages access to the
Logical Link Control (LLC): A protocol defined in the IEEE 802.2 communications channel (medium). It forms the lower layer of the IEEE
standard for data-link-level transmission control. It is the upper sublayer data link layer (OSI layer 2) which complements the Logical
of the IEEE Layer 2 (OSI) protocol that complements the MAC Link Control (LLC). MAC is a media-specific protocol within the
protocol. LLC is independent of any specific LAN technology. IEEE 802 specifications.
Multicast: An addressing mode in which a given frame is targeted to a Optical Time-Domain Reflectometry (OTDR): A method for
group of logically related stations. evaluating optical fiber based on detecting and measuring
backscattered (reflected) light. Used to measure fiber length and
Multicast Address: An address specifying a group of logically related attenuation, evaluate splice and connector joints, locate faults, and
stations on a network. Also called a group address. certify cabling systems.
Open Systems Interconnection (OSI): A communications reference
N model developed by the International Standards Organization (ISO) to
Network Interface Card (NIC): Also known as an adapter. An NIC is define all of the services a LAN should provide. This model defines
circuit board installed in a computing device to connect it to a network. seven layers, each of which provides a subset of all of the LAN
The NIC or adapter performs the hardware functions that are required to services. This layered approach allows small groups of related services
provide a computing device with physical communication capabilities. to be implemented in a modular fashion that makes designing network
software much more flexible.
Node: End point of a network connection. Nodes include any device
connected to a network such as file servers, printers, or workstations.
N-Way: Name originally used for the Ethernet Auto-Negotiation algorithm.
Spanning Tree: A structure that includes all the bridges and stations Synchronous Optical NETwork (SONET): Standardized by the
on an extended LAN in which there is never more than one active path American National Standards Institute (ANSI). A protocol for backbone
connecting any two stations. networks, capable of transmitting at extremely high speeds and
accommodating gigabit-level bandwidth.
Star Topology: A network configuration in which there is a central
point to which a group of systems are directly connected. All Time-Domain Reflectometry (TDR): A technique for measuring cable
transmissions from one system to another pass through this central lengths by timing the period between a test pulse and the reflection of
point. Ethernet 10Base-T is one example of a media system that uses the pulse from an impedance discontinuity on the cable. The returned
a star topology. All stations are connected through a central device waveform reveals many undesired cable conditions, including shorts,
called a hub. opens, and transmission anomalies due to excessive bends or
crushing. The length to any anomaly, including the unterminated cable
Station: A unique, addressable device on a network. A station is
end, may be computed from the relative time of the wave return and
identified by a destination address (DA).
nominal velocity of propagation of the pulse through the cable. See
Station Address: see MAC Address also Optical Time-Domain Reflectometry.
V W
Virtual Concatenation (VCAT): Process enabling transport pipes to be Wavelength-Division Multiplexing (WDM): Optical transmission
"right-sized" for various data payloads by allowing SONET/SDH technique that uses different light wavelengths to send data.
channels to be multiplexed in arbitrary arrangements. VCAT breaks Combination of two or more optical signals for transmission over a
down data packets and maps them into the base units of TDM frames. common optical path.
This data is then grouped in multiple data flows of varying size to create Wide-Area Network (WAN): A network that links data processing and
larger, aggregate payloads optimally sized to match available telecom equipment over a larger area than a single work site or
SONET/SDH pipe capacity. VCAT is applied at the end-points of the metropolitan area. A WAN usually links cities and is based on X.25
connections, which permits each channel used to be independently packet switching.
transmitted through a legacy transport network.
Virtual LAN: A method in which a port or set of ports in a bridge or 0-9
switch are grouped together and function as a single "virtual" LAN.
10 GigE: 10 Gigabit Ethernet
Virtual Private Network (VPN): One or more wide-area network links
over a shared public network, typically over the Internet or an IP 4B/5B Code: Scheme used to encode data for transmission in which
backbone from a network service provider (NSP), that simulates the 4-bit binary data values are encoded into 5-bit symbols for transmission
behavior of dedicated WAN links over leased lines. across the network media. 4B/5B is used with Ethernet 100Base-TX
and 100-Base-FX media systems.
Voice-over-Internet-Protocol (VoIP): Refers to communications
8B6T: Signal encoding method used with the 100Base-T4 Ethernet
services—voice, facsimile and/or voice-messaging applications—that
media system.
are transported via the Internet,rather than the public switched telephone
network. In an Internet-based telephone call, the voice signals are 8B/10B Code: Scheme used to encode data for transmission in which
converted to digital format and compressed/translated intoInternet 8-bit binary data values are encoded into 10-bit symbols for
protocol (IP) packets for transmission over the Internet; the process is transmission across the network media. 8B/10B is used with
reversed at the receiving end. 1000Base-X Gigabit Ethernet media systems and 10G Base-LX4.
LLC
Logical Link Control (see entry in glossary)
Acronyms Index
LSP
Label-Switched Path
LSR
Label-Switching Router
M
MAC
Medium Access Control (see entry in glossary)
MAN
Metropolitan-Area Network (see entry in glossary)
Mb/s
Megabits per second. One Mb/s equals one million bits per second.
MDI
9. Acronyms Index
A CoS Class of Service
ADM Add/Drop Multiplexer CRC Cyclic Redundancy Check (see entry in glossary)
APS Automatic Protection Switching CSMA/CD Carrier-Sense Multiple-Access with Collision Detect
(see entry in glossary)
ARCNET Attached Resource Computer Network
CRC Cyclic Redundancy Check (see entry in glossary)
ARP Address Resolution Protocol (see entry in glossary)
CWDM Coarse Wavelength-Division Multiplexing
ATM Asynchronous Transfer Mode (see entry in glossary)
(see entry in glossary)
B D
BER Bit Error Rate (see entry in glossary)
DIX Digital, Intel, and Xerox (the three companies that
BERT Bit-Error-Rate Testing released the original Ethernet specification in 1980)
b Bit (one binary digit) DUT Device Under Test
M O
MAC Medium Access Control (see entry in glossary) OAM&P Operation, Administration, Maintenance and Provisioning
MAN Metropolitan-Area Network (see entry in glossary) OTDR 1. Optical Time-Domain Reflectometry
Mb/s or Megabits per second. One Mb/s equals one million bits (see entry in glossary)
Mbps per second. 2. Optical Time-Domain Reflectometer
MDI Medium-Dependent Interface (see entry in glossary) OSI Open Systems Interconnection (see entry in glossary)
P
MII Media-Independent Interface (see entry in glossary)
PHB Per-Hop Behaviour
MEF Metro Ethernet Forum
PMD Polarization Mode Dispersion (see entry in glossary)
MEN Metropolitan Ethernet Network
PON Passive Optical Network (see entry in glossary)
MON Metropolitan Optical Network POTS Plain Old Telephone System
MPLS Multiprotocol Label Switching PRBS Pseudo-Random Binary Sequence
MSPP Multiservice Provisioning Platform PVC Permanent Virtual Circuit
R U
RJ Registered Jack (see entry in glossary) UTP Unshielded Twisted Pair (see entry in glossary)
RMON Remote Monitoring V
RPR Resilient Packet Ring VCAT Virtual Concatenation
S VLAN Virtual LAN (see entry in glossary)
SDH Synchronous Digital Hierarchy (see entry in glossary) VoIP Voice-over-Internet-Protocol (see entry in glossary)
SLA Service-Level Agreement VPLS Virtual Private LAN Service
SNMP Simple Network Management Protocol VPN Virtual Private Network (see entry in glossary)
SONET Synchronous Optical NETwork (see entry in glossary) VPSN Virtual Private Switched Network
STP Shielded Twisted Pair W
T WAN Wide-Area Network (see entry in glossary)
TDR Time-Domain Reflectometry (see entry in glossary) WDM Wavelength-Division Multiplexing (see entry in glossary)
TE Traffic Engineering
TLS Transparent LAN Service
ToS Type of Service
Notes
Acknowledgements
This guide would not have been possible without the collaboration and teamwork of EXFO staff; particularly, the technical expertise of
Mr. Bruno Giguère, B.Eng., Ethernet Product Manager; Mr. Matthew Demyttenaere, B.Eng., Application Engineer; Mr. Andre Leroux, M.Eng.,
Systems Engineer; and Mr. Scott Sumner, M.Eng., Strategic Marketing Manager.
ISBN 1-55342-004-7
Legal Deposit—National Library of Canada 2005
Legal Deposit—National Library of Quebec 2005
© 2005 EXFO Electro-Optical Engineering Inc., Quebec City, Canada. All Rights Reserved.
Ethernet Reference Guide: Your Everyday Ethernet Testing Reference Tool, 1st edition
Cover Ethernet.1AN: Cover Ethernet.1AN 5/7/07 10:13 AM Page 2