In a typical production environment, the CPUs on the machine are used simultaneously by the RAPCache database, RAPCache subscriber, and the operating system.
Sybase recommends that you set aside 50–70% of the CPU cores for the RAPCache database, depending on the number of activities that are executed in parallel. Consider the load on existing engines, resource contention, and overall response time when you determine whether to add or take engines offline. The RAPCache database contains performance monitoring utilities, such as the stored procedure sp_sysmon that you can use to measure load on existing engines, resource contention, overall response time, and to reveal the system's points of contention. See Adaptive Server Enterprise Performance and Tuning Series: Monitoring and Analyzing for Performance and the Adaptive Server Enterprise Performance and Tuning Series: Basics.
Sybase recommends that you allocate more engines to execution classes that are associated with CPU-intensive activities than those associated with less CPU-intensive activities. See Adaptive Server Enterprise Performance and Tuning Series: Basics.