Measuring Multiprocessor System Activity

Analyzing Performance on SMP Systems

You can monitor the activity of your SMP system by using the Performance console and its counters. To evaluate your system's performance, look at the following factors:

Processor utilization and queue length. You might need to partition the workload so that a particular processor handles a particular workload to achieve better performance. For more information, see "Optimizing and Tuning Multiprocessor Installations" later in this chapter.
Processor performance data, such as context switches, interrupts, threads, and processes. Activity rates and usage levels for these types of data that are higher than expected for a particular throughput can reveal inefficiencies in how your processors handle their workload. You might need to partition these types of activity. For more information, see "Optimizing and Tuning Multiprocessor Installations" later in this chapter.
Overall resource utilization on your system. Scaling to multiple processors can increase the load on resources such as disks, memory, and network components, and it might be necessary to increase the capacity of these components. For more information, see "SMP Impact on System Resources" later in this chapter.

Note

Application developers are in the best position to control how processes behave in an SMP environment. "Application Design and Multiprocessor Performance" later in this chapter provides guidelines for application developers who want to design programs that run well on multiprocessor systems.