SAN Choice: (2) Lefthand Networks NSM 2120 OR (1) EqualLogic PS5000X

[jeffswanson_biz]

Two Lefthand Networks NSM 2120 (clustered)

• 12 x 300GB 15k SAS Drives (each)

• 3.6TB RAW (each)

• 1.8TB usable, combined (RAID 10)

• Network RAID

• Cheaper to scale

OR

One EqualLogic PS5000X

• 16 x 400GB 10k SAS

• 6.4TB RAW

• 2.8TB useable (RAID 10)

• Dual controllers

• Great features

• Great performance

If price was the same, which would you choose and why? In your opinion, which product is more redundant and why?

I like LH because of the Network RAID feature and the ability to replicate to a VSA. To me the LH solution appears more redundant than a single tray EqualLogic product. Please educate me if I'm missing something.

[Anders_Gregerse]

I've been using Equallogic for some time now. No problems and management is a breeze. Installation and migration to new or more boxes is extremely easy and takes a few minutes of my time. They can be a bit expensive to expand if you just need more space, but not more bandwidth or more controllers.

Regarding redundancy it really depends on how you use your storage solution. On Equallogic you can have a LUN spanning several boxes and if you loose one of those boxes your LUN might be lost.

Also check how the future of Lefthand looks like, HP bought them and it is interesting how that developes.

[jeffswanson_biz]

If we went the EqualLogic route, I'm sure it would be sometime before we add a second tray. With EqaulLogic can you add more bandwidth or more controllers without adding storage?

[kingsfan01]

Hi Jeff,

You are going to get a variety of answers. My opinion is to go with the LeftHand solution. I have 2, 2-node clusters (3TB NSM 2060 SATA @ DR & 1.8TB DL320s/NSM 2120 SAS at the Data center).I have been extremely happy with our SAN. I've found that the ease of setup, use and total cost of implementation are huge selling points. I looked at some of the NetApp products before looking into LeftHand and found that a similar setup (Dual controller, total storage space, etc) on the NetApp came out almost twice what our LeftHand system cost. I'm sure part of that was NetApps a la carte pricing scheme and the aggressive discounts our reseller gave us on the LH equipment but money saved is money to spend elsewhere. I'm not too familiar with the EqualLogic equipment so I'll leave those comment for others who work with them.

In terms of redundancy... one of the other major selling points (for me) with the LH equipmentwas that I could split both nodes of my LH cluster between our two floors (seperate data centers) and provide floor to floor redundancy in the event of a power/HVAC/etc issue (not uncommon in our building). When the time comes to perform service on the cluster, I can take each node off-line individually without impacting any connected hosts. As for the network RAID functions, keep in mind also that with the LH equipment, with each additional node you don't just add additional space, but you increase your replication levels (should you decide to), increase your cluster wide cache/memory/processing power/etc and redundancy. It is a very cool system.

Have you tried downloading their VSA's yet? If you have a spare server or two (even a workstation would do) that has a nice chunk of DAS, you can setup the machines as ESXi boxes, run the VSAs from the ESXi box and setup a test cluster that responds the same way a physical cluster would (in terms of features/capabilities/etc. Obviously speed/redundancy is not the same).

Finally, you mentioned that the units are 3.6TB RAW, 1.8TB useable... are you sure about that? You should have a total cluster size of 3.6TB if you go network RAID 1. Why would you go RAID 10 on the box? We have ours setup as RAID 50 (from the factory) and it gives us our 1.8TB useable space (technically 1.6TB). You may want to check on the quote and if you really need to do RAID 10 on the box, with a 2-node cluster, you would probably be better off going RAID 50 on each unit. I have a feeling that they would come that way from the factory anyway though.

If you are interested, I have a spreadsheet detailing a variety of IOmeter speed tests I ran on our system when I first deployed it. Let me know and I'll shoot it over to you.

BTW - I do not work for LeftHand and am in no way associated with them other than being a customer... I just like telling people about my experience with them thus far because I find that there are few technology companies that are really worth boasting about.

Let me know if you have any questions or would like any additional information.

Thanks,

Tyler

[Paul_Lalonde]

I'd have to agree with Tyler. I've been an iSCSI advocate and user for many years now, and I'm sold on LeftHand's SAN/iQ solution. A few years ago, I would have recommended EqualLogic over anything else but I've come to recognize its limitations. The LeftHand solution scales better (in my opinion) and offers a higher degree of redundancy. Equallogic storage pools (collection of physical array members) have no inherent redundancy except at the physical chassis level. LeftHand's storage pools are bolstered by "network RAID" in addition to the physical chassis redundancy features.

If I were spending the money, I'd go with LeftHand and their NSM-2120 product. If you've taken a look at the new features offered by SAN/iQ version 8, you'll see a few compelling reasons to move in that direction as well!

Regards,

Paul

[jeffswanson_biz]

Thanks for the feedback Tyler and Paul, what's your take on HP aquiring Lefthand?

[Anders_Gregerse]

On Equallogic it's only possible to add bandwidth and controllers and storage at the same time, that is both the downside and upside depending on your needs.

[letoatrads]

I've RFP'ed both , and in the end I chose EQL for several reasons. At the time I made the choice ( about 2 years ago) , they were highly recommended by our VMWare rep. Yeah, they are supposed to be vendor agnostic as they work with all of them, but when we chatted 'off' the record he said 95% of the ISCSI in the South East region was EQL and for good reason.

The redundancy thing from my standpoint is KIS - Keep it Simple. EQL is a chassis, controller, and drives. Chassis is very simple and almost never gets crapped out ( dual PS) , Controllers should be dual as well, and drives...Raid 10 is good pretty good. The LH boxes I looked at were built on server chassis...motherboard, proc, ram and all the various single points of failure. This could have changed by now.

Ease of use - EQL was very very easy for me to config and use. I was used to a NetApp Fas240. LH interface was somewhat clunky.

My 2 cents, take it for what it is worth. I've used EQL for 3 large farms now and been nothing but happy. A primary point was EQL gave me an array for 10 days to test, LH wouldn't do the same. Again, may be different now as I know HP is better with the demo equipment than most.

[Paul_Lalonde]

Jeff, I've spoken at length with my regional LeftHand representatives and they're pretty unanimous on this one... HP will more or less "leave them alone" for the forseeable future. Eventually, SAN/iQ is expected to find its way into the MSA and EVA lines, but the core LeftHand products, sales and marketing programs, and support will stay the same.

Paul

[kingsfan01]

I agree with what Paul said. From everything I have heard from our LeftHand rep and our CDW rep, HP is going to position this inbetween the MSA and EVA lines and for the time being, doesn't have plans to touch SANi/Q. I need to expand my datacenter cluster and when I first heard rumors back at VMworld, I held off to see how things would play out but I feel that with the number of installations that LeftHand has in the market today and the fact that the HP management system (at least on the MSA side... I can't comment on the EVA) is nowhere near as robust as SANi/Q, they would probably leave it alone and have the LH group continue to develop/better the product. Similar to the way Microsoft buys products that are better than their own to bolster their offerings and get rid of the competition.

Tyler

[christianZ]

Iscsi comes with higher io latency than FC - and Lefthand storage seems to make it even worse IMHO (check it with "esxtop /d /F /j"). That would be a disadvantage for me.

They are not using the server's RAM as cache - only that one on the raid controllers. When you need a large capacity you have to get a lot of their boxes - always Raid 1 for redundancy. But it can be interesting for folks who wants to have sync mirror over 2 (or more) locations. However in that case the Datacore Sanmelody could be even better - they can run FC also, using internal RAM as cache and support servers with external storage - so you can get here more spindles per server - larger capacity less servers.

Just my thought.