Program execution times

Dynamic monitoring of IMS internal response times

IMS internal response times are not available in any dynamic report. However, many installations have written special transactions that always perform a fixed (usually small) number of DL/I calls, such as a GHU, ISRT, REPL, DLET sequence. By use of this transaction, an IMS MTO can dynamically monitor the responsiveness of the IMS system. This is not a precise measure, but an indication of what a user is experiencing at that time.

Daily monitoring of program execution

If the response-time breakdown data available from IMS PA or DFSILTA0 indicates large execution times, the cause might be related to one of the following:

• Program Load Time

  When daily monitoring indicates increases in execution times, the problem might be in the I/O subsystem.

• Program Initialization and Application Code

  After tuning, execution times are not likely to change significantly.

• DL/I Calls and I/O

  The IMS PA Internal Resource Usage report gives database I/O statistics. If these are high, the GTFPARS Data Set Overview report can be reviewed, together with the sizes and number of database buffers. Reorganize volatile databases regularly to reduce OSAM or VSAM overflow I/O. For further investigation of this area, the IMS Monitor should be run to obtain Call Summary and Region IWAIT reports.

• Other IMS Resource Delays

  The IMS Internal Resource Usage reports in IMS PA indicate if contention within program isolation activity is high, in which case, plans should be made to run and analyze the IMS PI trace.

• Dependent Region Modifications

  Examine the reasons for high elapsed time for dependent regions or for those application programs that exhibit excessive elapsed time for each transaction. You can use the IMS Monitor Program Summary report. Assuming that the dispatching priority is adequate, you can look for high schedule-to-first DL/I call elapsed times. Setting aside factors related to program load and paging, you need to find out if any modifications have been made in the dependent region code by your installation.

  If changes have been made to support accessing z/OS® data sets or the use of other z/OS services, you must evaluate the appropriateness and efficiency of these changes. One reason for these changes might be excessive overlay processing or initialization.

Detailed monitoring

GTFPARS Job Summary and Detail Trace reports for the IMS Message Regions are the best method of examining the system-related activities of the application program in detail. IMSASAP Program Trace reports provide detailed analysis of application activity related to database access in a DB/DC environment. IMS PA Database Trace reports show breakdowns of database update activity in a DB/DC environment. DFSPIRP0 should be used to investigate any long delays if PI lockout is suspected as a possible cause.

The time spent in database retrieval that incurs I/O can be a large contributor to program execution time. The frequencies of each type of DL/I call to each database segment type and the frequencies of IMS IWAITS are given in the Call Summary report. If database-related delays are suspected as the cause of high program-execution times, that report helps identify the particular call and database combination that is responsible. The I/O wait times are shown in the IMS Monitor Region IWAIT report.

IWAIT times greater than 100 milliseconds are usually caused by interference from shared DASD or over commitment of resources. Databases using VSAM often experience shorter IWAITs, because the I/O buffer control managed by VSAM is able to meet the request without physical I/O. Excessive chaining of records to overflow tracks can be a contributing factor. Sometimes the data transfer rate of the storage device itself has a limiting effect.

In addition to database design considerations, the two principal strategies for decreasing I/O IWAIT time are:

• Reduce the frequency of I/Os by adding more buffers. This entails identifying inadequate IMS buffer allocation.
• Examine the I/O resource usage to discover if the distribution of data sets is a problem. This is a more protracted analysis that uses detailed output obtained from GTF and RMF™ traces.

A more extensive action is to change the DL/I structure or database organization.

Related reading: For more information on changing the DL/I structure or changing database organization, see IMS Version 13 Database Administration.