IMS message processing in DB/DC and DCCTL environments
This topic identifies possible performance problems associated with message processing and the message queues on an IMS™ system.
Dynamic monitoring
MFS and Queue Pool statistics can indicate problems in this area. Pool sizes cannot be dynamically increased. They must be increased at the next IMS start. Normally, once this aspect of performance is tuned, it requires only periodic monitoring.
Daily monitoring
If IMS output message processing is a problem area, IMS input message processing is probably affected by the same delays. Investigate any large or variable output queue time of IMS message queue and MFS (message format buffer pool) statistics using the IMS PA Resource Usage reports and DC Queue Manager Trace reports.
If output queue times are high in a DB/DC environment, check the IMS PA DC Queue Manager Trace report for output message lengths and run the IMS Monitor to study the Communications IWAIT report.
Detailed monitoring
- Message Queue Processing
- IMS PA can be used to investigate
the IMS Queue Manager in detail,
using the DC Queue Manager Trace report. DFSUTR20 and IMSASAP give
statistics on frequency and duration of communications-related IWAITs.
GTFPARS Job Summary and Detail Trace reports for the IMS control region can be used to examine the performance of MFS and Message Queue I/O.
- Message Queue Data Set Tuning
- The three data sets, IMS.QBLKS, IMS.SHMSG, and IMS.LGMSG, contain the directory and the short
and long message segments for the message queues. In a busy online
system with some constraints on the terminal I/O and conventional
SPA pools, these data sets are the most active and critical. Place
these data sets on different devices to minimize seek time. Check
to see if the I/O activity is balanced between short messages and
long messages. You might have to allow for a possible difference in
device access times.
You can do this by examining the IMS Monitor Communication IWAIT report. The entries are organized by line number and reveal high frequency or high I/O elapsed times. Each line in the report identifies the DD name causing the IWAIT, so that you can pick out those that pertain to IMS.QBLKS, IMS.LGMSG, and IMS.SHMSG. The Region IWAIT report similarly identifies message queue I/O and totals IWAIT times for the region.
You can define up to 10 long and 10 short message queues.
Related reading: For more information on this option, see IMS Version 13 System Definition.
- MFS Usage
- The GTFPARS Job Summary report for the IMS control region provides data on EXCP-SIO-IO timings to FORMATA/B. The IMS Monitor indicates in the Communication IWAIT report any IWAITs for Format Pool Space or for Directory or Block loading.
Message format library optimization
The active IMS.FORMATA/B data set is the source of all MFS format blocks used during online execution and is usually a busy data set.
You can check the frequency or high I/O elapsed times against this data set using the IMS Monitor Communication IWAIT report. Report lines identified by I/O=DIR=fn or I/O=BLK=fn, where fn is the format block name, show I/O IWAIT time for directory lookup and block fetch by increasing the size of the format buffers. This pool size is specified with the FORMAT keyword on the BUFPOOLS macro for system definition. You can override the value for the online execution using the FBP parameter in the EXEC statement.
To reduce I/O to the directory, you can also create an index of track addresses of the blocks (or only those frequently referenced) and keep this index present in the MFS pool. The index is built when you execute the MFSRVC procedure, specifying INDEX on the utility control statement. You also include the names of the format or message blocks to be in the index.
Multiple reads of the active library data set occur when the size of the MFS control blocks stored on DASD is greater than either the track size or the z/OS® block size.