How can we measure response time for the pc_server?

• Article Type: General
• Product: Aleph
• Product Version: 18.01

Description:
How can we measure response time for the pc_server?

Resolution:
util s/4/2 PC Session Statistics shows you how many sessions there were in a particular time period, but it doesn't show the response time.

Version 18 rep_ver 12277 added an "ELAPSED-TIME" line to the pc_server log file for each transaction. The line format is:

ELAPSED-TIME: <SERVICE> <CURRENT TIME> <ELAPSED TIME IN SECONDS>

The following grep can be used to create a file (jerrytime) with these lines in it:

> grep ELAPSED-TIME pc_server_6991.log > jerrytime

The file created looks like this:

ELAPSED-TIME: C1001 15:57:51 0.0800
ELAPSED-TIME: C1205 15:57:51 0.0937
ELAPSED-TIME: C1001 15:57:51 0.0802
ELAPSED-TIME: C1205 15:57:52 0.0743
ELAPSED-TIME: C1205 15:57:52 0.0768
ELAPSED-TIME: C1205 15:57:52 0.0699
ELAPSED-TIME: C0104 15:57:52 0.0404
ELAPSED-TIME: C0442 15:57:52 1.1075

Then the following grep's can show how many transactions took a particular length of time:

grep -c ' 0.' jerrytime (this gives you the number less than 1 second)
grep -c ' 1.' jerrytime (this gives you the number 1.0 - 1.9 seconds)
grep -c ' 2.' jerrytime (this gives you the number 2.0 - 2.9 seconds)
grep -c ' 3.' jerrytime (this gives you the number 3.0 - 3.9 seconds)
<etc.>
grep -c ' 10.' jerrytime (this gives you the number 10 - 10.9 seconds)
grep -c ' 11.' jerrytime (this gives you the number 11 - 11.9 seconds)
grep -c ' 12.' jerrytime (this gives you the number 12 - 12.9 seconds)
<etc.>

What can you do with this?

1. Compare the results from day to day to see if there are particular days/times when transactions are slow.

2. See if particular programs (the "Cnnnn" value) account for the longer transactions

3. "view" the log and examine the longer cases in detail, using the following command, for example, to locate them:

/ 12.

• Article last edited: 10/8/2013