HPE Storage Users Group

We have had some ongoing performance issues with some of our SQL servers that we are running as VMware Virtual machines. We have played with queue depths in VMware, Windows, etc. etc. We have the drives on question running on nearline drives which I realize that the nearlines are not made for performance but the LUNS we have presented are huge at 16TB each as they are used for warehousing data. So, at this point, I am thinking we are trying to squeeze more out the NLs than they can give. I saw numbers in another post on this board rating the NLs I think at 75 IOPS each, 10K FC at 150 IOPS and 15K FC at 200 IOPS. In the management console I have seen the Nearlines running at 200 IOPS or greater for a considerable length of time. We currently have 3 CPGS

NL - 72 3 TB drives
FC - 96 450 GB Drives
32 900 GB Drives
SSD - 32 100 GB Drives
16 400 GB Drives

These are SQL servers running on VMware so looking for any guidance with what some other folks have experienced as far as performance numbers.

Thanks
Mike

You are definitely running your NL drives way too hot. Even the 75 IOPs per drive is for "acceptable" response time. When I design something with NL drives I try to scope 0 IOPs on the NL drives (meaning no planned performance requirement). At most 50 IOPs should be the goal. At 200 your response time must stink.

You need to find the offenders and move them off to other drives or get something like Adaptive Optimization to move the busy bits to faster drives. If you have System Reporter or 3.1.2 and later there are AO reports available that will tell you if you have a good IO density to use AO.

NL's running at 200 iops can be possible but only if they are processing 100% sequential iops. (e.g. AO or a rebuild/sparing of a failed drive). Please have a look at the servicetimes during this load. if below 20msec i wouldn't worry to much, but if they are more like e.g. 50msec, than it is clear they are heavily overloaded.

are you using AO?