topic Re: Quick Performance+Reliability Comparison between the P4000 VSA & the P4500 in StoreVirtual Storage

Quick Performance+Reliability Comparison between the P4000 VSA & the P4500

WFPL4E — Tue, 27 Dec 2011 17:54:56 GMT

Hello,

Does anybody have some rough comparison figures on Performance and Reliability (suppose on equal hardware) between the P4000 VSA (Virtual SAN appliance) and the P4500, if the total needed storage capacity will not exceed 10 TB?

I understand that the P4500 can scale out much more than the software-based VSA, however, this is not needed here.

The environment will involve:

Primary Site (A): where 2 Lefthand SANs will be replicating

Secondary Site (B): whose existence is just to play the role of a fail-over Site.

I appreciate any valuable comment

Thank you,

Re: Quick Performance+Reliability Comparison between the P4000 VSA & the P4500

Paul Hutchings — Thu, 29 Dec 2011 15:42:15 GMT

I've heard anecdotal evidence from people running P4000 that they have seen 90-95% of the performance they get from physical nodes from their VSA's.

I would suggest that if you go with a good hardware config i.e. hardware RAID5 or RAID10 and a good RAID controller with lots of BBWC that you will be good.

The VSA can, I believe, scale just as much as the P4300/P4500 in terms of nodes, however you "only" have a single ethernet connection.

I don't know if that's limited to 1gbps or if you can use 10gbps if you have it in your ESXi host though - perhaps someone does?

I'd be very interested in knowing the answer as we have 10 "spare" VSA licenses that I haven't yet needed to use, but that's a lot of storage I have access to should the need arise.

Re: Quick Performance+Reliability Comparison between the P4000 VSA & the P4500

WFPL4E — Fri, 30 Dec 2011 04:18:33 GMT

- Thank you Paul.

- Per my partner's vision, I may most likely go with a P4500 because it involves more hardware which - despite the lack of enough scientific evidence - seems to him more reliable than a mere software abstraction layer.

- I would be curious and interested to know though about the experience of an environment that has been running for a while with multi-site VSA deployment that capitalizes on the features of the LeftHand SAN

I appreciate any input which may contribute to my decision to adopt VSA in future deployments.

Thanks again

Re: Quick Performance+Reliability Comparison between the P4000 VSA & the P4500

Paul Hutchings — Fri, 30 Dec 2011 09:57:46 GMT

It depends on what you're using it for I suspect. Sure, the virtualisation layer will add some overhead, but there's a big difference between, say, high IOPS highly random database apps vs. lower general file serving access.

The beauty of the VSA is that you can download it and try it of course, only problem is you need the hardware spare to throw at it :)

Re: Quick Performance+Reliability Comparison between the P4000 VSA & the P4500

M.Braak — Mon, 02 Jan 2012 22:15:23 GMT

Use VSA only for very small workloads. A vsa can only have 1 vCPU and is exhausted very quickly which limites throughput..

Re: Quick Performance+Reliability Comparison between the P4000 VSA & the P4500

oikjn — Tue, 03 Jan 2012 00:06:32 GMT

I"m not sure if its "supported", but I know for a fact you can run each VSA with at least two virtual processors (at least in Hyper-V).

I"m seeing on average 55% CPU usage with an average of 350IOPS and 40MBps on each of a two node VSA cluster running raid 10 running off of a software raid 1 onboard card using two 1TB 7200RPM SATAII Hdds each. The CPU is an intel xeon e5630 running at 2.5ghz. I'm a little concerned about the CPU usage, but the VSA seems to run pretty well to me. I can't wait until the prices from the floods drop back down and I can throw some more spindles and some real hardware raid on the back end of the VSAs.

Re: Quick Performance+Reliability Comparison between the P4000 VSA & the P4500

WFPL4E — Tue, 03 Jan 2012 02:11:16 GMT

I appreciate the Input from every body. You are right about the Spindle

Hyper-V consumes a lot more CPU cycles as compared to ESXi.

However, 55% is a lot to put at stake.

Thx

Re: Quick Performance+Reliability Comparison between the P4000 VSA & the P4500

oikjn — Tue, 03 Jan 2012 02:33:36 GMT

yea... I think it will scale a lot more. While the CMC reports 55%, M$'s VMM only reports ~28% and the host certanly isn't CPU bound and I think that there is more room for IOPS than there might seem with my raw initial numbers.

I see spikes of abour 1800IOPS and 140MBps which I assume are when the hyper-v uses some sort of caching on the host to bood HDD performance.

I plan on putting the final system on two supermicro servers with an LSI raid cards w/ their cachecade2.0 feature to use a dozen slow HDDs + SSDs to give me the performance I need.

That said, I'm a small install with only ~1200IOPS where I could probably get away with DAS but for the need for shared storage and the need for remote replication.