March 3, 2010 at 5:08 pm
Hi,
We have SQL Server 2005 EE x64 with SP3 and all the drives are from SAN & all the drives are RAID 10. I would like to calculate Average Disk Seconds/Read & Average Disk Seconds/write to know disk performance.
What will be good/accepted threshold values for:
Average Disk Seconds/Read
Average Disk Seconds/write
Is there any special method to calculate the above values if the drives are from SAN & RAID 10?
thanks
March 3, 2010 at 5:18 pm
Check out this article for some tips
http://msdn.microsoft.com/en-us/library/cc966412.aspx
Each environment will be different. You should test your system, baseline it, and then trend it over time.
Jason...AKA CirqueDeSQLeil
_______________________________________________
I have given a name to my pain...MCM SQL Server, MVP
SQL RNNR
Posting Performance Based Questions - Gail Shaw[/url]
Learn Extended Events
March 4, 2010 at 2:26 am
For sec/read and sec/write, under 10ms is good. 10-20ms is average. Over 50ms is bad.
Gail Shaw
Microsoft Certified Master: SQL Server, MVP, M.Sc (Comp Sci)
SQL In The Wild: Discussions on DB performance with occasional diversions into recoverability
March 4, 2010 at 3:39 pm
Thanks,
I have multiplied the PhysicalDisk(1 D:)\Avg. Disk sec/Read & Avg. Disk sec/Write by 1000 to get the value in ms and most of the values are with in 15 ms -
:-):-)
but what should the accepted value for Avg. Disk Queue Length?
what are the units for Avg. Disk Queue Length
do I need to multiply Avg. Disk Queue Length value with 1000 similar to Disk sec/Read & Disk sec/Write?
I'm getting the below values for Avg. Disk Queue Length for the Backup drive. So do I need to multiply these valuse by 1000? or I need to take the values as it is from perfmon counters?
0.213329237
0.206662699
0.179996544
23.79954305
0.166663467
0.173330005
0.179996544
0.186663083
0.166663467
please advice
March 4, 2010 at 3:59 pm
gmamata7 (3/4/2010)
but what should the accepted value for Avg. Disk Queue Length?
No way to say. Depends completely on your IO subsystem and, if you've got a SAN, that counter is near-impossible to interpret.
Gail Shaw
Microsoft Certified Master: SQL Server, MVP, M.Sc (Comp Sci)
SQL In The Wild: Discussions on DB performance with occasional diversions into recoverability
March 5, 2010 at 12:05 pm
thanks Gail,
No way to say. Depends completely on your IO subsystem and, if you've got a SAN, that counter is near-impossible to interpret
Is there a way to find the correct values for Avg.Disk Queue length from SAN side for a particular disk (in my case, for the Z drive (backup drive))?
thanks
March 5, 2010 at 12:09 pm
There's too much between the OS and the disks to get a meaningful threshold for queue length when there's a SAN. It's one of those counter where a value far above average for your server is bad, but that's all that can really be said.
The podcast mentioned in this post may be of interest
http://sqlinthewild.co.za/index.php/2009/02/14/do-you-know-how-to-use-perfmon/
Gail Shaw
Microsoft Certified Master: SQL Server, MVP, M.Sc (Comp Sci)
SQL In The Wild: Discussions on DB performance with occasional diversions into recoverability
March 5, 2010 at 12:21 pm
thank you,
If the storage is from SAN, the readings from Perfmon for the below counters will be the correct values unlike Avg Disk Queue length?
Average Disk Seconds/Read
Average Disk Seconds/write
thanks
March 5, 2010 at 12:34 pm
I never said the avg disk queue length was incorrect. It's not. All perfmon counters report accurate, correct values.
The avg queue length is near-impossible to interpret on a SAN. There's no value that's good or bad, you have to compare with what's normal for that server and drive. You cannot, with a SAN, look at a single value of queue length and say if it's good or bad.
Gail Shaw
Microsoft Certified Master: SQL Server, MVP, M.Sc (Comp Sci)
SQL In The Wild: Discussions on DB performance with occasional diversions into recoverability
March 5, 2010 at 12:39 pm
I think, as per microsoft, Disk Queue length should not exceed 0.02 seconds
March 5, 2010 at 12:44 pm
thank you,
I understand that perfmon gives accurate values but when we have SAN it's near impossible to interpret correct values for Avg.Disk Queue length.
Question:
Is this applicable to Avg. Disk sec/Read & Avg. Disk sec/Write counters too. i.e near impossible to interpret correct values from SAN?
thanks
March 5, 2010 at 1:01 pm
we also have all our sqls running on SANs....
I've been using perfmon, quest Perf Analysis, Spotlight and most importantly - HP SE (SANs monitoring\management soft)
they are all reporting the same perf values for the counters you spcified.
March 5, 2010 at 1:43 pm
SD1999 (3/5/2010)
I think, as per microsoft, Disk Queue length should not exceed 0.02 seconds
Queue measures the number of things waiting. Hence it's an integer value and it's a count of things, not a time. You're thinking of the sec/read or sec/write that should not exceed 0.02 seconds.
Gail Shaw
Microsoft Certified Master: SQL Server, MVP, M.Sc (Comp Sci)
SQL In The Wild: Discussions on DB performance with occasional diversions into recoverability
March 5, 2010 at 1:46 pm
gmamata7 (3/5/2010)
Is this applicable to Avg. Disk sec/Read & Avg. Disk sec/Write counters too. i.e near impossible to interpret correct values from SAN?
If it was, do you think I would have given you the values that I gave in my first post in this thread?
I'm guessing that you haven't listened to the podcast that I recommended. I do advice that you do so.
Gail Shaw
Microsoft Certified Master: SQL Server, MVP, M.Sc (Comp Sci)
SQL In The Wild: Discussions on DB performance with occasional diversions into recoverability
March 5, 2010 at 2:07 pm
Queue measures the number of things waiting. Hence it's an integer value and it's a count of things, not a time. You're thinking of the sec/read or sec/write that should not exceed 0.02 seconds.[/quote]
Oops, you are absolutely right. I was talking about latency for sec\read , sec\write
Viewing 15 posts - 1 through 14 (of 14 total)
You must be logged in to reply to this topic. Login to reply