Performance degradation due to global enqueues when using GRS Ring
After you upgrade from a single CICS region to multiple CICS regions, you experience severe performance degradation due to global enqueues. You are using a global resource serialization (GRS) Ring. Your application issues a temporary storage enqueue an update, and a dequeue. The CICS trace shows about 2 seconds between the EXEC CICS ENQUEUE and the following MVS enqueue.
The amount of CPU being used by the regions actually decreases. The regions are not waiting on CPU and different tasks are receiving CPU time.
GRS Star might be needed in large complexes.
Resolving the problem
Use the GRS Star configuration with the coupling facility (CF). See the z/OS MVS Planning: Global Resource Serialization guide under Methods of Serializing Global Resources for more details on which method to choose.
In this particular case, CA Technologies Multi-Image Manager (MIM) was being used with the GRS Ring. When GRS Star was used in production everything ran approximately 25-30% faster than without global enqueues and 300-400% faster than with GRS Ring.
CICS/TS CICS TS CICS Transaction Server
Translate this page: