Analyzing disk performance in Windows NT and 2000 using Microsoft's built-in reporting tools is harder than it needs to be because some of the information is confusing. Specifically, the % Disk Time, % Disk Read Time and % Disk Write Time numbers produced by the PhysicalDisk Object don't mean what they seem to -- or what Microsoft's documentation implies they mean.
The first clue that something odd is happening is that sometimes the numbers don't add up. That is % Disk Read Time and % Disk Write Time add to more than the % Disk Time. The second clue is that the read and write numbers can add to more than 100 percent, which implies the disk is busy more than all the time.
While that would be a nice piece of performance enhancement, that's not what's actually going on. Despite what the associated "explain" text implies, the % Disk Time isn't measuring the time the disk is actually active. The number is derived from the time between the time I/O operations start and they finish -- including the time the I/O request spends waiting in the queue before it is processed. If there is significant queuing, the actual figure can be well over 100 percent. Since this would be extremely confusing, Microsoft initially "fixed" the problem by arbitrarily capping % Disk Time at 100 percent, hence the confusing numbers -- and the apparent poor performance of disks when there are a lot of I/Os in the queue.
Mark Friedman, author of the "Windows 2000 Performance Guide" from
Rick Cook has been writing about mass storage since the days when the term meant an 80K floppy disk. The computers he learned on used ferrite cores and magnetic drums. For the last twenty years he has been a freelance writer specializing in storage and other computer issues.
This was first published in September 2002