Highlighted
jjgrinwis
2 Iron

unisphere analyzer pool vs raidgroup stats

Jump to solution

What happend to the SP Cache stats for luns that are build on pools.

A customer is using pools, no fast cache license yet, but we can't find any SP Cache stats for the individual luns.

If we look at the classic raidgroup luns, the SP Cache stats can be found so we can check for any write cache misses etc etc.

Regards,

John

Labels (2)
0 Kudos
1 Solution

Accepted Solutions
kelleg
5 Tungsten

Re: unisphere analyzer pool vs raidgroup stats

Jump to solution

Because Pools do not allow you to change the cache settings, you can not see any data about them - the cache is handled by the Pool drivers and not the individual LUNs within the pools.

glen

View solution in original post

0 Kudos
8 Replies
kelleg
5 Tungsten

Re: unisphere analyzer pool vs raidgroup stats

Jump to solution

Because Pools do not allow you to change the cache settings, you can not see any data about them - the cache is handled by the Pool drivers and not the individual LUNs within the pools.

glen

View solution in original post

0 Kudos
jjgrinwis
2 Iron

Re: unisphere analyzer pool vs raidgroup stats

Jump to solution

ok, so if we really want details stats from a lun we still need to build it on the "classic" raidgroups.

Beside cache stat's I'm also not seeing any stats about disk-crossings and I can't imagine that with a misaligned FS on a pool we will never suffer from disk-crossings.Would be nice to see some more stats from pool luns.

John

0 Kudos
kelleg
5 Tungsten

Re: unisphere analyzer pool vs raidgroup stats

Jump to solution

Remember that disk-crossing are not a problem with file-system alignment, that's stripe-crossings. Besides, it's always considered best practice to align the file system (except on Wondows 2008).

glen

0 Kudos
jjgrinwis
2 Iron

Re: unisphere analyzer pool vs raidgroup stats

Jump to solution

We where able to solve a misalligned fs on a lun used for mssql by looking at the crossings a couple of months ago.

Thing is that would have taken us some more time to solve, if this had been build on a pool lun, because of the missings stats.

John

0 Kudos
SKT2
4 Tellurium

Re: unisphere analyzer pool vs raidgroup stats

Jump to solution

what you meant by "We where able to solve a misalligned fs on a lun used for mssql by looking at the crossings a couple of months ago".. as Mentioned earlier disk crossings and stripe crossings are 2 different things.. Can you explain your theory a little more so we know what may have happened in your case.

0 Kudos
jjgrinwis
2 Iron

Re: unisphere analyzer pool vs raidgroup stats

Jump to solution

Below two screenshots from analyzer showing the diskcrossings:

Disk crossings% before changing block size:

Unknown.jpeg

Disk crossings% after changing block size

Unknown.png

disk crossings dropped to a max of 32%, causing a lot less read IOP/s.

My problem is with the pools, how can we ever generate stats like this to troubleshoot performance issue's when using pools.

0 Kudos
SKT2
4 Tellurium

Re: unisphere analyzer pool vs raidgroup stats

Jump to solution

Thanks for posting this buddy. What was the block size(default) earlier? And what is the new block size?

But the NTFS allocation size(block size) is different than the partition offset. They are two different things.

Can you also check what is your parition offset

0 Kudos
jjgrinwis
2 Iron

Re: unisphere analyzer pool vs raidgroup stats

Jump to solution

SQL was runing on 2008, disk was aligned.

For some reason they used 48K block size, we've changed that to 64K, the recommended value which solved the above problem.

0 Kudos