Planning for performance in a shared-queues environment

In a shared-queues environment, individual transaction performance might be poorer than in a non-shared-queues environment. However, because of increased parallelism and workload balancing, overall system performance and throughput should increase. And because input transactions are processed by the originating local IMS system (if a message region is available), you can influence IMS performance by controlling the number of dependent regions for each IMS system.

Each of the following can have an effect on system performance:

• IMS execution parameters: QBUF=, QBUFMAX=, LGMSGSZ=, and SHMSGSZ=
• CQS Local Structure Definition parameter, SYSCHKPT=
• CQS Global Structure Definition parameters: OBJAVGSZ=, OVFLWMAX=, and STRMIN=
• Size of the shared queues structures on the coupling facility
• Size of the z/OS® system log structure on the coupling facility
• Number of z/OS system log data sets to back up the z/OS system log structure
• Frequency of structure checkpoints

The size of all log records has increased by eight bytes, and the size of Queue Manager and EMH log records has increased by an additional 32 bytes.

The SPA pool is no longer used, decreasing the amount of data logged for each checkpoint, and increasing the amount of available storage. The SPA is now maintained with its input and output messages.

For Fast Path systems, using the Fast Path Input Edit Router (DBFHAGU0) should be considered carefully because messages defined as Local Only will have priority over all other messages (Local First and Global Only).