How to Investigate Disk Bottleneck

On previous entry, i explained how to determine that you have enough ram on your server. Today, i will tell you about how to examine disk bottleneck. For this purpose, you need monitor a few performance counters :

• PhysicalDisk : Average Read Queue Length
• PhysicalDisk : Average Write Queue Length
• PhysicalDisk : Disk Transfer / sec
• PhysicalDisk : Idle Time
• PhysicalDisk : % Disk Time

I admit that, it is usually difficult to understand exact meaning of counters. But anyway, i will try to explain with the values what they should be.

Average Read Queue Length indicates how many system request are waiting for disk access. This value should be less than 2. 

Average Write Queue Length indicates average number of write requests that are queued. This value should be less than 2. 

Disk Transfer / sec shows the rate or read and write operations on disk. If it is greater than 25 disk I/O per second, you have poor response time. This value of the counter should be below 25.

Idle Time indicates idle time of your disk. Normally, value of the counter should be greater than %20. 

Disk Time is the percentage of elapsed time that the selected disk drive was busy servicing read or write requests. If it is more than %50, it indicates a disk bottleneck 

Beside these counters, you can check following counters to determine whether it's a bottleneck.

Average disk seconds/read and average disk seconds/write need to be below around 12ms.