Page 525 - Handbook of Modern Telecommunications

4-56 CRC Handbook of Modern Telecommunications, Second Edition

RT1: Time up to the first character of the response on the screen of the user
RT2: Time up to the last character of the response on the screen of the user
User time Intranet time User time

Response time (1) = RT1

Response time (2) = RT2

FIGu RE 4.4.1 Response-time definitions.

An additional issue for SLA management is to determine whether an event affecting the service at the
SAP is causing a complete service outage (service fully unavailable) or a partial service outage (degraded
service available).
Critical success factors for measuring and computing availability are:
• Trouble tickets
• Proactive tools
• Workforce with dispatch capabilities
• Efficient help desk
• Skill and experience of subject matter experts
Response time is one of the key metrics in all SLAs. Its definition varies, but usually users consider the
period between sending the inquiry and receiving the full answer as response time. Figure 4.4.1 displays
the differences between two response-time (RT) alternatives:
• Time up to the first character of the response on the screen of the user
• Time up to the last character of the response on the screen of the user

The second definition is a better suited definition for the working cycle of users. The difference
between RT2 and RT1 depends on many factors, such as the throughput of the backbone and access net-
works, servers in these networks, number of hops, and the hardware/software capabilities of the client’s
workstation or browser. Present measurement technology offers the following alternatives:
• Monitors and packet analyzers: These analyzers filter and interpret packets and draw inferences
about application response times based on these results. These monitors are passively listening to
the network traffic and calculate the time it takes specific packets to get from source to destina-
tion. They can read the content of packages revealing eventual application errors and inefficiency,
but they cannot measure response time end to end.
• Synthetic workload tools: These tools issue live traffic to get a consistent measurement of response
time on a particular connection in the intranet or for a given application. These tools are installed
on servers, desktops, or both. They typically send Transmission Control Protocol (TCP) messages
or Structured Query Language (SQL) queries to servers and measure the time of the reply. Results
from multiple sources are correlated to give a more detailed view of intranet response times. They
are very accurate with respect to the end-to-end response time.
• Application agents: These agents work within or alongside applications, using software that moni-
tors keystrokes and commands to track down how long a specific transaction takes. They can run
on both the client and server. They clock specific portions of the application at the server or at the
workstation. The use of agents needs customization and the correlation of many measurements in
order to give users a performance estimate about their intranet.
• Use of application response-time measurement (ARM) management information bases (MIBs):
ARM defines application programming interfaces (APIs) that allow programmers to write agents