The Effect of Solid State Disks on Processor Usage

I am now encountering servers and client computers that have solid state disks (SSDs). My experience is showing that this sometimes causes improper analysis of performance measures through Performance Monitor (perfmon.exe) and Resource Monitor (accessible through the Resource Monitor button on the Performance table of Task Manager).

In the past, poor interactive response time of a system, or complete “locking up” of a system was not necessarily accompanied by high processor usage, but high processor usage was usually indicated that a system was slow to respond or locked up. Disk IO is usually the primary performance bottleneck in a system because disks are the component that includes mechanical limitations of physical disk access. Response time can be viewed in the Disk Activity panel of the Disk tab of Resource Monitor, or measured by Avg. Disk sec/Read and Avg. Disk sec/Write in Performance Monitor. Be sure to get the “seconds” in the numerator when you select counters—there are several counters that are very similar.

When we use SSDs, this changes. The biggest bottleneck is reduced or removed, and the computer can process much more information per second, and therefore processor usage goes up—often to 100%. I have seen systems that were either running at 0% or 100% with no in between. After eliminating all other problems, I began to realize that once all other potential problems were eliminated, that this high processor usage was good rather than bad, and was seldom accompanied by the “locking up” of a system—other activities could run or continue to run and receive a reasonable share of “time slices” of a system.

The takeaway is that if you have a system with SSDs and there is 100% processor usage but things seem to be working fine, everything is probably OK and is likely to be a desirable status.

0 replies

Leave a Reply

Want to join the discussion?
Feel free to contribute!

Leave a Reply

Your email address will not be published. Required fields are marked *