Mike Maddox wrote to us to take exception to the advice offered in Metron's Tips Archive regarding the analysis of performance of Raid 5 disks on NT systems. In response to the advice given in Windows Performance Tip15 Mike says:

"The only way to tell for sure how busy the array is, is to calculate the number of physical I/Os per second per disk. Multiply the number of writes per second by 4 and add the number of reads per second. Then divide that total by the number of disks in the array.

If the number of I/Os per second per drive is at or above the manufacturer's recommended I/O rate, the disk array is likely a bottleneck. If much less, and especially if this is the busiest array on the machine, the cache may be a bottleneck.

If the cache is a bottleneck, 30-40% may or may not be a good guideline. Some disk arrays may saturate at less than that if the cache is small and record sizes large, while larger caches with very fast drives may saturate only under extreme conditions.

System manufacturers like Compaq do offer documents about how to analyze their I/O performance. I strongly recommend them to anyone who wants understand I/O performance thoroughly."

Mike acknowledges that different Operating Systems may vary in the detail of the performance data they report. But Windows NT/2000 is a known quantity.