stripe size and block size

[dcoz]

Hi guys,

I have been reading quite a bit about vscsistats etc. And am keen to hear other peoples views on what sizes they place on their SAN for stripe sizes to try and gain better performace.

Correct me if i am wrong but, if you watch a vm from vscsistats and you find it is using 8K blocks most of the time it is better to have the stripe size on the array the same to match this.

So if i have multiple vms all reading and writting to a vmfs volume, all using slightly different block size writes how can i optimise the stipe size?

Am i thinking about this to much?

I the gain so small it si not worth it except very specific workloads?

Any thoughts or corrects would be welcomed.

Thanks

Dougie

[4nd7]

Hi Dougie,

I think you should let the stripe size of the array as the vendor recommends (EMC Clariion, 64k, Netapp 4k) and format the partitions of the vms for the specific workloads that they will manage.

Thanks,

Dan