Is it possible to measure the load(%) on disks in a raid5

[Yaop]

I can see theserver CPU is 25% stressed, it's using a quater of the memory but I can't see how much stress the harddrives is being put through? - Please help!

 

[SomeJoe7777]

Performance for disks is not as clear-cut as a % load. However, there are ways you can get an idea of how much you're stressing your RAID array.

This test is applicable to server environments, where several requests are being made to the array at once. The test doesn't mean much if only 1 user or process is accessing the array at a time, like would be the case on an enthusiasts/gamers machine.

This test works in Windows XP and Server 2003. Probably works in Vista, but I make no guarantees.

1. Right-click "My Computer", go to "Manage".
2. On the left, expand "System Tools", then expand "Performance Logs and Alerts".
3. Click on "Counter Logs".
4. On the right, right-click and choose "New Log Settings".
5. Name your log "Disk Queue Length".
6. A dialog box will pop up asking you to configure the performance log. Click the "Add Counters" button.
7. Choose "Select counters from computer", and choose your computer's name from the drop-down list.
8. Under "Performance object", select "Logical Disk".
9. Choose "Select counters from list", and click on "Avg. Disk Queue Length".
10. Choose "Select instances from list", and click on the drive letter that is associated with your RAID-5 array.
11. Click the "Add" button, then click "Close".
12. Under the "Sample data every" section, choose 1 second as the interval.
13. Go to the "Log Files" tab, under "Log file type" select "Text File (Comma Delimited)".
14. Click the "Configure" button, click "Browse" and select a location to save the log files. Click "OK".
15. Go to the "Schedule" tab. Under both "Start log" and "Stop log" sections, click "Manually (using the shortcut menu)".
16. Click the "OK" button at the bottom.

You now have a new performance log defined in the right-hand pane of Computer Management.

17. When you're ready to test, click the "Disk Queue Length" performance log, and right-click it and select "Start" (or click the Play button in the toolbar). This will start logging.

18. Do whatever test you want that stresses your RAID array.

19. Right-click the "Disk Queue Length" performance log and click stop to stop logging.
20. You now have a log file containing 1-second samples of the Disk Queue length in the folder/file you specified in step 14.

You can now import this log file data into Excel and graph it if you want to, or just look at it in Notepad or some other text editor.

In general, the rule of thumb is that the maximum average disk queue length for an array should not exceed certain values, as follows:

A. If your array is using desktop-rated drives, the maximum average disk queue length should not exceed 1 per disk in the array. For example, if you have a RAID array of 4 desktop-rated drives, the maximum average disk queue length as reported in this log file should not exceed 4 when the array was under the maximum workload.

B. If your array is using enterprise-rated drives (10K RPM SATA/SCSI, or 15K RPM SCSI), the maximum average disk queue length should not exceed 2 per disk in the array. For example, if you have a RAID array of 4 15K RPM SCSI enterprise-rated drives, the maximum average disk queue length as reported in this log file should not exceed 8 when the array was under the maximum workload.

If your disk load is such that these queue lengths are being exceeded routinely, you have 2 choices:

1. Add disks to the array to spread the workload out amongst more spindles.

2. Change the drives in the array to drives that can handle a higher workload. Drives that are rated to have the highest IOPs under queue depths of 2-4 will improve the array performance.

 

[Guest]

wow you're pretty handy to have around