Memory usage requirements - USS

This topic describes the memory usage considerations that apply to the operation of the CAE Server on USS.

Typically, when running the CAE Server on USS, memory usage requirements are:
  • A 500MB Java™ heap
  • A REGION that is 300MB greater than the Java heap size
  • A setting of _BPX_SHAREAS=NO

The garbage collector in the mainframe Java Virtual Machine (JVM) allows the memory usage to approach the existing Java Heap size before garbage collecting. The JVM attempts to grow the heap when it is needed for even one-time memory demands.

If you set the heap size to 600, one-time large memory demands can cause the JVM to get close to that size even if a smaller heap size would have been sufficient overall. Once the JVM reaches the max heap size it will tend to stay close to that heap size even if it doesn't need to because garbage collection is not initiated until the current heap is full.

Refer to Chapter 2, "Understanding the Garbage Collector" from the latest edition of the IBM Developer Kit and Runtime Environment, Java 2 Technology Edition Diagnostics Guide (SC34-6358) for more details.

The following situations could produce an increased demand for memory:

  • An increased number of alerts
  • A higher number of users logged on at the same time
  • Larger tables displayed in the client

Large viewing requests

If a user makes a large viewing request (for example, to view many thousands of SQL Text summaries), the request will be refused if either the CAE Server or CAE Agent has insufficient memory to satisfy the request. If this occurs, a message appears in the data table indicating that the CAE Server or CAE Agent refused the request.

If the CAE Agent refused the request, or if you are running the CAE Server on USS, you must increase the CQM_HEAP and REGION and restart the appropriate address space.

If the restriction is in the CAE Server and the CAE Server is not running on USS, contact IBM® Software Support for help in increasing the heap size for the CAE Server.

Note: For large SQL text summary displays we recommend you take advantage of the new TopN feature rather than increasing the heap size. Also note that a viewing request can run short of memory resources if your browsing history still contains previous large viewing requests. You should try using the Clear History button before deciding that there are insufficient resources for your request.

Alerts and memory usage

Alerts are intended to be used sparingly. If alerts arrive on the message board at a rate faster than you would like to receive emails, then you probably have the alerts threshold set too low. Alerts are not intended as a means of storing data that you go back to and review at a later time; exceptions serve that purpose.

Optimizing Event AutoClear

You can set Event AutoClear to a shorter period if needed. This is set by going to the MITs panel in the Profiles & Configurations browser and selecting Event in the tree and clicking on the link embedded in the appropriate text, for example:

"Automatically remove messages about this event from the message board 9 days after the initial posting time."