HP LH4r HP Netserver & Microsoft Terminal Server 4.0 - Page 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. Percent Average %Processor Time 100 90 80 70 60 50 40 30 20 10 0 LH 3 0.05 Text/0.5 Menu Delay 35.23 0.05 Text/2.00 Menu Delay 6.65 Delay Lengths Figure 18 0.5 Text/4.00 Menu Delay 1.45 Hewlett-Packard Company 23