

11. WORKLOAD CHARACTERIZATION › 11.6 Case Studies › 11.6.3 IMS Transaction
11.6.3 IMS Transaction
To investigate potential IMS transaction classes, we used
Workload Characterization to analyze DETAIL IMS transaction
data collected with the optional _USRXISU exit in the IMS
Component. In particular, we analyzed the IMS transaction
data to determine patterns based on CPU time, message queue
calls, and database calls.
To identify potential IMS transaction classes, we analyzed
the DETAIL timespan IMSISU File. (The IMSISU File is not
normally supported in the DETAIL timespan, but was included
through the use of an optional exit.)
We specified three data elements on the Workload
Characterization screen:
ISUCPUTM - The CPU time consumed by the transaction
ISUDBTOT - The total number of database calls issued by
the transaction
ISUMSTOT - The total number of message queue calls issued
by the transaction
We selected four data elements as reporting variables:
ISUDBGU - The number of database GET UNIQUEs issued by
the transaction
ISUDBGN - The number of database GET NEXTs issued by
the transaction
ISUMSGET - The number of message queue GET UNIQUE and GET
NEXTs issued by the transaction. This value
is computed from ISUMSGGU and ISUMSGGN.
ISUCPUTM - The CPU time consumed by the transaction
The centroids of the six clusters that we developed are shown
in Figure 11-33. Cluster 2 represents 71% of the
transactions that were processed. We could potentially use
the transactions in this cluster to define an IMS transaction
class. However, we would have to perform further analysis on
the ALL_OBS SAS data set to verify that the DBD requirements
of the transactions do not conflict (ART78, LEE83).
CLUSTER CENTROIDS
CLUSTER ISUCPUTM ISUDBTOT ISUMSTOT RADIUS NO ASSIGN PCT
------- -------- -------- -------- ------ --------- ---
1 0.13 11.9 1.23 1.00 219 11
2 0.21 9.8 2.99 1.00 1423 71
3 0.17 17.8 0.5 1.00 77 4
4 1.13 150.3 1.0 2.00 41 2
5 0.85 1.3 61.2 1.50 137 7
6 0.53 0.9 0.1 1.00 103 5
Figure 11-33. IMS Cluster Centroids Report
Copyright © 2014 CA.
All rights reserved.
 
|
|