Note that the EQ systems are almost always dual controller where each LH node is a cascade of single points of failure. Comparing what happens when an EQ fails to what happens when an LH node fails is therefore invalid as the LH is much more likely to fail than BOTH controllers on the EQ are. A single controller failure in the EQ is a non event.

So I’ve read a few horror stories with LH here (and on other blogs/forums around the web). Any similar stories on EQ? I haven’t seen much thus far.

I believe EQ supports proper multipathing for VMWare environments since 4.x (one of the updates).

If you want to fail over to the async replicated array, not sure that is supported or possible (think cache coherency).

w.r.t dedicated switches vs shared switches, only if your shared switches can do real QoS (need real hardware queues and lots of buffering) on a per VLAN/port basis.

The decision pretty much boils down to LH vs EQ (Starwind VSA on R710s is a distant consideration).

Reading this thread from beginning to end caused me to flip back and forth a few times.

We’re looking at an EQ PS4100XV with async replication to PS4100E versus a P4300 starter SAN kit (2 x DL185 in active/active mode).

After reading all the above, I really like the EQ solution but I’m leaning towards LH. There is about a 20% price difference and LH’s active/active cluster w/ virtual IP is really what we wanted from the get go (high availability without fuss). If my main EQ node goes down, there’s a lot of work to repoint ESX hosts to the secondary (lower performing) node whereas there’s nothing to do in LH.

Any additional comment and/or insight will be appreciated.
Also, what are your thoughts on dedicated switches vs sharing an existing L3, GigE switch using VLAN?

We’ve similar problems over the years with companies HP had purchased that almost match your story identically. The latest being IBRIX. But in the past it has taken them about 2 to 3 years to come back to previous support levels and expertise. Here’s hoping LH and IBRIX both get there eventually.

Originally EQL was my preferred choice since before the Dell takeover. Back in 2008 when we were looking at buying our first iSCSI SAN We were hoping Dell’s EQL aquisition might drive down their prices a bit, but this turned out not to be the case, They were just a bit over budget for us at the time.

LH came in a fair bit cheaper for the solution they were offering us back then. The RAW 4.5 TB in a pair of NSM2060′s with 6 x 450GB 15k SAS per node was similar to that availble in the EQL solution. I should add that we were naive in looking at RAW capacity and not sufficiently taking into account the overhead of network RAID 1.

LH also sold us on the concept of their scalable ‘pay as you grow’ approach (grow your storage, Processor, IOPS and Network IO at the same time as you add further nodes in future expansions). We also liked the ability to do network RAID 1 on a per volume basis so your important volumes could stay up even if a node fails.

After almost 3 years running LH, its become clear to us that the TCO for the LH solution is really very high, and the majority of points in the sales pitch have proved very misleading.

Didnt take us long to figure out that RAW capacity figures of a LH cluster are pretty meaningless. We overestimated the reliability of the individulal nodes. We were thinking we could run some volumes without network RAID to maximise space. But really you cant. You really have to run network RAID 1 for everything, because these boxes can – and do – go down, for reasons ranging from hardware failures, to SAN-IQ software bugs. Thus the maximum usable storage with LH should always be in seen in reality as 50% of the RAW capacity at best. The £/TB cost suddenly starts to look a lot less attractive. Your storage costs just doubled.

then the HP takeover happened. We had some big fears about that. Assurances were given to us from LH. they turned out to be pretty worthless. I’ve never been a particularly big fan of HP, and wouldnt willingly be an HP customer. The old LH account management team gave us assurances that LH would remain pretty independent and we shouldnt worry. but:

- the Dell hardware options got canned (we are a dell shop, so this was a big deal)
- support portal went away. we lost access to all the tech info, patches, downloads. HP’s wrote to us telling us that we should start using the ITRC but the lefthand information just wasnt available. It was a complete nightmare. All self help resources just gone, overnight.
- result of the above was we had to uplift our support from 8x5xNBD (which was actually sufficient at the time) to 24x7x365, because we had to call someone if we wanted the smallest bit of info that we used to be able to get from the web.
- when we tried to do this, we found prices for support renewals and uplifts had gone through the roof (50% higher than they were before the HP takeover)
- Where once we had a a UK toll free number to call for support and could always get through to a friendly knowledgable engineer in the US who would resolve whatever issue or queries we had on that first call. This number was suddenly withdrawn and and 1st line UK tech support started going through India. They barely knew what the products were. They also worked on a callback basis. the callbacks would usually come back 3h 50 minutes, so as to just stay inside the SLA, but questions would never be answered on these initial callbacks. The calls would invariably get escalated to the same team in the US, the difference being that issues that took an hour or so to resolve now took several days just to get to the same person we could at one time call direct. A real insult considering we were now expected to pay 50% more for the support contract.
- No account management. nobody in HP would take our calls, listen to out complaints. We couldnt get any product or roadmap information. To all intents and purposes LH had ceased to exist, and we were left on our own.

Fast forward to now, the ITRC is a bit better and we at least have an account manager again, but actual tech support has got even worse. recently we’ve had some serious issues post SAN-IQ 8.5 where solution pack bits started acting up in our environment. Responses from the US team who weere once so good is now totally woeful. Things like ‘try rebooting, uninstalling reinstalling’ We had snapshopts buiding up and up for days because of VSS writer failures. They didnt want to get involved in any root cause analysis. We sent them logs, they wouldnt look at them. They were more concerned in blaming some pretty random stuff in our environment. One tech even asked us to swap out all the patch cables in use in our SAN. We had to turn off VSS writers in the end. HP couldnt find the answer and their closing statement was “we’ve done all we can, we cant help you anymore, we dont know. Sorry”.

Despite all this, we recently we wanted to expand our SAN. We wanted to add more nodes to our cluster and begin to realise the benefits of increased IO, throughput and disk that LH promised. We also wanted to take advantage of the new network RAID 5 level and to try to claw back some usable space in the cluster. (Network RAID 5 currently requires a minimum of 4 nodes in a cluster to make it work)

HP originally lead us to believe that we could add nodes to an existing cluster in pairs – even if they were not the same size as our original ones. They told us that provided we added equal sized nodes in pairs then we could scale out the cluster and get all the benefits. (We questioned that on a couple of occasions and actually have it in writing from them confirming it).

We were about to buy a a pair of P4500 7.2TB nodes when we discovered accidentally from a LH tech, that the information we’d been given by pre-sales was false. We’d have to put these new nodes in a NEW cluster if we wanted to use all the storage. This really means starting a new SAN. If put these new nodes in our current cluster, we’d loose about 60% of their capacity. They actually tried to tell us this wasnt a big deal!

When we bought the 2.2TB NSM 2060′s in 2008 they were at the sweet spot of cost / vs capacity for us. 3 years later, The P4500 7.2TB nodes represent the same optimal choice for us in terms of £/TB. But to grow our SAN the way LH sold it, we would have to continue to buy much smaller boxes at less optimal £/TB, or spend more money to replace the whole lot with the newer nodes. Neither of these choices were optimal, and made a mockery of LH’s ‘pay as you grow’ marketing

Additionally, HP couldnt offer us any hardware that would allow us to expand our current cluster without some wasted space as no ‘exact match’ for the Dell based 2.2TB NSM2060′s exists with HP hardware. So we’d still end up paying for ‘wasted’ space.

as a customer, we had a choice to make. exapnd the current cluster with lower capacity more expensive nodes, with wasted space, eating up rack space and power in the process..
…or, start a new 2 node cluster (really a seperate SAN) with the new 7.2TB capacity nodes.

It seems you can expand your space, OR you can expand your throughout and IO, but you cant do both as the same time at a reasonable price point. Expansion of a LH SAN is always going to mean compromise in this respect, its never going to fulfil its pre-sales promise.

best we’d get for our money was to buy the 7.2TB nodes, create a new cluster, accept the management overhead of managing 2 pools of storage, and sacrifice the scale-up of CPU, IOPS and Network IO and still have no access to the newer network RAID5 features, as both clusters would have 2 nodes each. We also realised that this story would be repeated at the next stage of expansion.

We’ve really regretted choosing LH over EQL back in 2008. It was a bit more expensive initially, but with all the weakneses of the LH solution that have become evident over time, coupled with issues caused by the HP aquisition we’d have been much better off in the long run.

As it turns out, thats what we’re doing. We’ve re-evaluated the EQL kit, and we’ve decided not to buy LH for this expansion. Instead we’re migrating the whole SAN to EQL.

Hope this helps anyone out there about to make a similar choice

@ roland

You state that LH is only for SMB, Test and non-mission critical infrastructures. Did you ever work with LH ? Your opionion is not that objective if you didn’t work with it. If you have a 2 node setup with network raid 10 over 2 sites. Isn’t that mission critical HA? If the backplane of the EQL blows? What happens than?