HP Netserver & Microsoft Terminal Server 4.0

Hewlett-Packard Company 23
Affect of Delays (“Think” Time) on Processor Utilization
The Visual Test script used for this exercise was specifically designed to help
characterize, using a limited number of client sessions, multiprocessor utilization on
HP NetServers running TSE.
When trying to use this data to estimate the actual number of client sessions a server
can support, it is very important to remember that even minor changes in the length
of the delay between keystrokes will significantly affect processor utilization. And
consequently, it will affect the number of sessions that a server can support.
In order to illustrate the effect of delay (i.e., “think”) time on processor utilization, the
24 client configuration was run on the LH 3 three times using different delay settings.
The following chart shows how processor utilization is affected when the keystroke
delay is increased from 0.05 seconds between text keystrokes and 2.0 seconds
between keystrokes that initiate menu operations to (1) 0.05 seconds and 0.5 seconds
and (2) 0.5 second and 4.0 seconds.
NOTE While 4 seconds is a rather significant delay when using a word
processor, using an 4-second delay to characterize the “think” time
between bank transactions may not be unrealistic.
With the exception of the runs described in this section, all tests were
performed with delays of 0.05 seconds between text keystrokes and
2.0 seconds between keystrokes that initiate menu operations.
Average %Processor Time
0
10
20
30
40
50
60
70
80
90
100
Delay Lengths
Percent
LH 3
35.23 6.65 1.45
0.05 Text/0.5 Menu Delay 0.05 Text/2.00 Menu Delay 0.5 Text/4.00 Menu Delay
Figure 18