topic Re: EVA disk group best practice in HPE EVA Storage

EVA disk group best practice

zeno75 — Sun, 17 Jun 2012 12:19:13 GMT

We've have some performance issues with our EVA, based on the HP best practice document it recommends to have as many disks in a single disk group which we have done and on it were running VMware, SQL and the ocupancy is 128 disks with 95% capacity utilization.

What we've found is frequently were getting some high disk latency on the VMware datastores and also SQL get timeout errors. Due to disk requirements we've purchased some new disks and need to deceide to create a new DG or add the disks to the existing DG.

Given the current circumstances can anyone provide some suggestions - the vendor recommended we create a new DG with the 60+ disks b/c we won't get much performance gain by having the extra disks in the current DG.

Secondly whats the best practice in adding disks to an existing DG eg. if we have a DG with 100+ disks and we need to add another 60 new, do we add the disks one at a time or all 60 to the DG at once.

Cheers

Re: EVA disk group best practice

Johan Guldmyr — Mon, 18 Jun 2012 08:54:49 GMT

How many disks do you have at the moment? ( EVAs like the number of disks to be divisible by 8)

Which EVA model do you have?

HP support has a nice performance analysis tool. There is also apparently also something similar in CV these days. Perhaps this is something you could use - to create a baseline of your current performance and after you have added the new disks you could then see if there are improvements or not.

Why wouldn't adding disks to a DG help?

I suppose one benefit of creating a new disk group is that you could dedicate those disks to the services or servers that have vdisks in that disk group.

One quite common thing when it comes to EVA performance is to have the controllers unevenly utilized. evaperf can help with monitoring this. And the solution is to spread the vdisks between the controllers.

Re: EVA disk group best practice

zeno75 — Tue, 19 Jun 2012 08:32:20 GMT

We were told by the SI with documentation from HP that there's a diminishing return on performance once you get past 120 disks in the DG right now were at 120, and increasing the 60 or so disks we only get much improvement in performance.

We are alos reviewing the balacing of vDisks between the SP's.

One side question were using Window 2008 R2 and I with MPIO and the HP DSM - however I'm trying to find information on what I should configured for the "Preferred Path/Mode" on the EVA. For Vmware we have "Path X/Failover/Failback" but for Win2k8 R2 I can't seem to find any documentation on whats the best configuration.

Re: EVA disk group best practice

Johan Guldmyr — Tue, 19 Jun 2012 11:32:17 GMT

If you don't have Failback, then if for some reason the vdisks are moved to another controller you'll have to divide the vdisks again.

Re: EVA disk group best practice

zeno75 — Tue, 19 Jun 2012 14:13:54 GMT

So does that mean we should be using "Failover/Failback" configuration? Johan

Re: EVA disk group best practice

Johan Guldmyr — Tue, 19 Jun 2012 14:23:25 GMT

As far as I know Windows don't mind if the path to a LUN changes back and forth.

Other OS might be different.