August 27, 2010 at 3:29 am
Hi all
on a current logical volume in perfmon which is using a scalre of 100.00 my average reading is coming back as 1.860 which would indicate a value of 186 for he avg. disk queue length correct?
If this is so then its extremely high and is maximum reaches the thousands.
Other counters such as buffer cache hit averages 99.5% and Avg. Disk sec/read 0.013 and Avg. Disk sec/write 0.015 on a scale of 1000.00 which would indicate it to be 13ms and 15ms respectively which seems a reasonable speed.
In that case what could be causing the disk queue to be so high?
Thanks for any steer in the right direction.
August 27, 2010 at 3:52 am
Hi,
Actually the scale is only relevant when looking at the graph. So if you're looking at the graph and the value is 52 (for example) you should divide by the scale (not multiply).
Either way though, if you're actually looking at the values provided rather than the graph, those are the true values. So, if your average reading is 1.860 and you got that from the values shown then that's your queue length (divided by the number of disks in your raid configuration of course).
Hope that helps.
August 27, 2010 at 3:52 am
I also got the same kind of problem in one of my dev server,But no resolution till yet 😀
when do we get high spikes for "disk queue length".NO patch/SP/hardware change on this server
-------Bhuvnesh----------
I work only to learn Sql Server...though my company pays me for getting their stuff done;-)
August 27, 2010 at 5:25 am
Hi Karl - That was a useful explanation thank you.
August 27, 2010 at 8:24 am
Are you actually experiencing IO bottlenecks etc? A disk queue length of 1.86 doesn't indicate a problem on its own. A disk queue length > 2 (per spindle / indivual physical dislk) would suggest a bottleneck, although occasional spike are to be expected.
So, for example, a 2 disk RAID1 drive showing an Avg Queue Length of > 4 would be an issue, a 5 disk RAID5 showing an Avg Disk Queue Length > 10 would be an issue etc. Also check % Disk Time (>50%), and look for certain *IO* WAITYPES on processes to indicate Disk IO issues.
September 1, 2010 at 3:16 am
Just one more addition to this..... when we select the counters we've an option to select the individual drive (i.e. C or D etc) this will give more specific info if you run the counter by selecting your data or log drive rather than selecting the total one..... which will be the average of all the drives....
Rohit
Viewing 6 posts - 1 through 5 (of 5 total)
You must be logged in to reply to this topic. Login to reply