CPU usage with BSR

In this example, the system uses BMC SUPEROPTIMIZER® products for every CICS and IMS region, and every user logs on to CICS and IMS through a session manager. In that environment, the SUPEROPTIMIZER products consume CPU under the address spaces of CICS and IMS. VTAM consumes CPU under the address spaces of CICS, IMS, and the session manager when any of those applications issues a VTAM SEND or RECEIVE.

When the SUPEROPTIMIZER products are replaced by ULTRAOPT by using BSR, the VTAM CPU time used will tend to drop; and the Optimizer time that is used in CICS and IMS is almost entirely removed (because all sessions use BSR).

ULTRAOPT consumes CPU time under the session manager; so its CPU time rises. However, that increase in the session manager should be offset by the decreases in CICS and IMS CPU time. Message throughput might increase, because ULTRAOPT moves the Optimizer workload away from CICS and IMS subsystem TCB and performs work by dispatching SRBs instead.

The total number of bytes that are reduced should be greater under ULTRAOPT because it can optimize READ BUFFER commands that are generated by the session manager during a session switch.

If the session manager is used to access other applications that were previously not optimized, the optimization benefits of those newly optimized sessions result in some additional CPU cost.

To accurately measure the overall CPU savings of BSR, normalize the results by dividing the total CPU time that is expended across all regions by the total number of optimized messages. The result is the CPU expense per optimized message. It is necessary for a fair assessment because ULTRAOPT might result in higher throughput and more load on the subsystems per unit of real time. The CPU time should also be divided by the total number of bytes that are reduced under both test cases to determine the CPU cost saved per byte.