Information Technology Reference
In-Depth Information
Catalog browse
12
Max response time = 10 s
10
8
6
4
2
0
1
50
100
150
200
Number of round trip transactions
Figure 9.11
Round trip performance for peak catalog browse workload after performance correc-
tions have been applied
The empty system performance of the log on/log off code has not been affected
by the code changes. This response pattern is expected because the correction is
supposed to make a difference only when other transaction groups are added to the
workload mix. The purchase steps code and catalog browse code both show slightly
improved performance in empty system workloads. This is a pleasant surprise be-
cause the code corrections were focused on cross-transaction performance degrada-
tion due to a poor common utility-routine-sharing strategy. The improved sharing
strategy also helped each transaction group independently of the workload mix.
Because the individual transaction group workload rerun responses stay well
below the requirement maximums, our fi nal step for this workload plan is to rerun
the transaction group mixes to verify that the performance fi xes did speed up the
transactions in competition for computing resources. In Figure 9.12 we see the fi rst
results of our transaction mix performance test reruns.
Log on
Purchase
8
8
Purch max
7 s
7
7
6
6
5
5
4
4
Log on max
3 s
3
3
2
2
1
1
1
50
100
150
200
250
300
350
400
450
500
Number of round trip transactions
Figure 9.12
Round trip performance for peak logon + purchase workload mix after performance
corrections have been applied
Search WWH ::
Custom Search