Disk performance monitoring

[Guest TPGBrennan]

The description of the Physical disk\current disk queue length perfmon

counter indicates it should be no more than 1.5 to 2 times the number of

spindles that make up the disk. Quite the argument has broken out among my

co-workers, two SAn vendors, and myself over how to interpret this. does

this mean that if we have 48 spindles in a disk group this counter can be as

high as 98 without indicating a problem?

[Guest Dave Patrick]

Re: Disk performance monitoring

 

Maybe these help.

 

http://technet2.microsoft.com/windowsserver/en/library/1b669468-2dc4-462e-881f-14842f54751a1033.mspx?mfr=true

http://technet2.microsoft.com/windowsserver/en/library/2e8550f7-c4c1-4a0c-82e4-74655a6116411033.mspx?mfr=true

 

or try asking them here. microsoft.public.storage

 

http://www.microsoft.com/communities/newsgroups/en-us/default.aspx?dg=microsoft.public.storage

 

 

--

 

Regards,

 

Dave Patrick ....Please no email replies - reply in newsgroup.

Microsoft Certified Professional

Microsoft MVP [Windows]

http://www.microsoft.com/protect

 

"TPGBrennan" wrote:

> The description of the Physical disk\current disk queue length perfmon

> counter indicates it should be no more than 1.5 to 2 times the number of

> spindles that make up the disk. Quite the argument has broken out among

> my

> co-workers, two SAn vendors, and myself over how to interpret this. does

> this mean that if we have 48 spindles in a disk group this counter can be

> as

> high as 98 without indicating a problem?

[Guest Leythos]

Re: Disk performance monitoring

 

In article <A33C6434-A21F-4877-B515-7766EF61F39D@microsoft.com>,

TPGBrennan@discussions.microsoft.com says...

> The description of the Physical disk\current disk queue length perfmon

> counter indicates it should be no more than 1.5 to 2 times the number of

> spindles that make up the disk. Quite the argument has broken out among my

> co-workers, two SAn vendors, and myself over how to interpret this. does

> this mean that if we have 48 spindles in a disk group this counter can be as

> high as 98 without indicating a problem?

 

Generic rule of thumb, from real world experience, as soon as the READ

or WRITE AVG Queue runs about 80% of the number of DRIVES, not spindles

you have in an array, and it maintains that average, you are going to

suffer performance.

 

I have a small server with 8 drives in a RAID-5, all 15K SAS drives,

that is a SQL 2000 Database (just the MDF files) and the queue runs READ

AVG runs 0.033 to 1.890, WRITE runs 0.001 to 1.474 AVG, under load they

can both spike to 11 or 12 for a single shot, lasting 1 or 2 MS, but

that's it.

 

On another server, an imagining system, the vendor setup 3xSAS 15K

drives and it shows a read queue of 8.x to 290.x average and write runs

about the same - the server is trashed, it has a serious hardware

problem that causes this - a new server shows < 2.0 for the same exact

config of OS/Software.

 

Your caching can also have a BIG impact on performance, get as much as

you can (Memory on the RAID card).

 

So, with 200+ servers with RAID installed, my experience says "DRIVES"

and to be less than the AVG of all drives in the array, in fact, if

you're running more than 80% most of the time then you need more drives.

 

--

 

Leythos

- Igitur qui desiderat pacem, praeparet bellum.

- Calling an illegal alien an "undocumented worker" is like calling a

drug dealer an "unlicensed pharmacist"

spam999free@rrohio.com (remove 999 for proper email address)