Without an adequate number of engines (CPU resources), tasks and worker processes must wait for access to Adaptive Server engines, and response time can be slow. Many factors determine the number of engines needed by the system, such as whether the query is CPU intensive or I/O intensive, or, at different times, both:
Worker processes tend to spend time waiting for disk I/O and other system resources while other tasks are active on the CPU.
Queries that perform sorts and aggregates tend to be more CPU-intensive.
Execution classes and engine affinity bindings on parallel CPU-intensive queries can have complex effects on the system. If there are not enough CPUs, performance for both serial and parallel queries, can be degraded.
See Chapter 5, “Distributing Engine Resources,” in the book Performance and Tuning: Basics for more information.
