Вы находитесь на странице: 1из 2

© 2001 Giga Information Group

Copyright and Material Usage Guidelines

November 29, 2001

SLA Benchmark Metrics


Adria Ferguson
Contributing Analyst: Colin Rankine

Catalyst
A client inquiry

Question
What are some benchmarks we should use for service-level agreements in terms of system availability and
break/fix for servers, mainframes, help desk, peripherals and related equipment?

Answer
Many service-level agreement (SLA) benchmarks that Giga reviews tend to focus on current and target SLAs
(for six or 12 months out  showing improved performance after six and 12 months). Most SLAs break up
availability targets by equipment, application and function. For example, the SLA would stipulate specific
target levels for the major applications for the specific equipment, and for the type of environment, such as a
development environment vs. a production environment or a mission-critical application vs. a noncritical
application. Additionally, many SLAs will create availability rates for both prime time and nonprime time
operations. Most SLAs will also provide time to respond rates and time to resolution rates for various degrees
of problems (see table below).

Problem Management (example)


Description Current Service Six-Month Target 12-Month Target
Level Service Level Service Level
Time to Respond
*Severity 1 95% within 15 min. 98% within 15 min. 99.5% within 15 min
Severity 2 95% within 2 hrs. 97% within 2 hrs. 98.5% within 2 hrs.
Severity 3 95% within 4 hrs. 96% within 4 hrs. 98% within 4 hrs.
Severity 4 95% within 2 days 96% within 2 days 98% within 2 days
Time to Resolve
*Severity 1 90% within 4 hrs. 90% within 2 hrs. 95% within 2 hrs.
Severity 2 90% within 24 hrs. 90% within 4 hrs. 95% within 4 hrs.
Severity 3 90% within 3 days 90% within 24 hrs. 95% within 24 hrs.
Severity 4 90% within 5 days 90% within 3 days 95% within 3 days

Note: Severity 1 is for mission-critical applications


Source: Giga Information Group

For business or mission-critical applications in midrange distributed server environments: Time to respond
usually ranges from one day to within 15 minutes (most are within four to six hours), and time to resolve
usually ranges from within four days to within four hours (most are within 24 hours). In SLAs, the majority
of availability rates for production environments range from 99.9 percent to 99.95 percent, and the majority

IdeaByte ♦ SLA Benchmark Metrics


RIB-112001-00196
© 2001 Giga Information Group
All rights reserved. Reproduction or redistribution in any form without the prior permission of Giga Information Group is expressly prohibited. This information
is provided on an “as is” basis and without express or implied warranties. Although this information is believed to be accurate at the time of publication, Giga
Information Group cannot and does not warrant the accuracy, completeness or suitability of this information or that the information is correct.
SLA Benchmark Metrics ♦ Adria Ferguson

of availability rates for development environments range from 97.0 percent to 99.5 percent. Additionally,
nonpeak time usually has about a 1 percent lower availability rate than peak time for the same applications.
Regarding help desk SLAs, first call to resolution is often within 70 percent, speed of answer is often within
30 seconds, and an abandon rate of less than 8 percent is usually satisfactory. For peripheral equipment, a
bundle print request within 15 minutes and a bundle print retrieval within one hour are typical.

For mainframe applications: The primary efficiency metric for mainframe data center operations is
MIPS/FTE (the FTE number counts everything except for applications groups). The current number is 22
MIPS/FTE. Doing the same for nonmainframe server operations is more difficult, with so much depending on
the server, OS type, standardization, etc. In terms of service levels, it depends on the application specifics, but
for well-run mainframe applications, 99.99 percent scheduled availability is attainable, and we do see 99.995
percent as an objective occasionally. Most organizations still have at least a few hours of scheduled downtime
per month. Other metrics commonly measured are batch abend rate (<0.5 percent or <0.25 percent is an
aggressive objective), percentage of console operator tasks automated (over 95 percent is easily attainable),
percentage of tape mounts automated, average outage duration (<20 minutes is a recovery time objective),
batch and work order completion, user wait time for help desk, and percentage of problems resolved on the
first call to the help desk.

IdeaByte ♦ RIB-112001-00196 ♦ www.gigaweb.com


© 2001 Giga Information Group
Page 2 of 2

Вам также может понравиться