Why I chose Dell EqualLogic and not HP Lefthand
Posted on May 13, 2010
I spent quite a lot of time weighing up the pros and cons of the two most-similar disk arrays on the market, HP Lefthand and Dell EqualLogic. With similar names (PS4000E – Dell; P4000 – HP) and similar initially observable features (that is, they’re both lots of disks in a box, they’re both iSCSI) it was quite difficult to work out whether one was better than the other or not.
I chose Dell EqualLogic in the end, which means the rest of this post is likely to be flavoured with a bit of Dell bias, but I’ll try and explain why I went that way (depsite at the time receiving quite poor sales support due to placing the order in the middle of some bizzare change from partner channels to Dell Direct).
I should take a moment out here to say that within the storage market there’s been a new trend, headed by Dell and HP (well, technically headed by EqualLogic and Lefthand before they were respectively bought out). Traditional (read: older) storage arrays worked on a controller and disk tray model, you bought a controller and disk trays, you could buy more disk trays for more space but eventually the controller would get too busy and you’d need another of those and so on. This is often explained using a train as an analogy (with the engine as the controller and carriages as disk trays).
The newer arrays come all in the same box. You have controllers enough for the disks that are in the box and they’re bundled together, to increase capacity you buy another “controler+disk” box and plug it in, they scale much more efficiently as a result because there’s never a situation where a controller is overworked.
Before I did anything the guys at Dell helped me out by suggesting that we do some checks on the company’s current IOPS – I needed to see what the existing servers were doing in terms of disk accesses to make sure that any specified solution would be able to meet the same level (and offer some headroom for expansion) – having done this Dell extracted a 95% percentile score for the company of about 900 IOPS – this was the yard stick for solutions.
The four high-level reasons that I chose Dell over HP were:
- EqualLogic arrays seemed to offer higher performance
- EqualLogic solution was actually cheaper
- Anecdotal evidence suggested EqualLogic arrays were of a better build quality than LeftHand
- General advice from multiple vendors and independent experts hinted that EqualLogic had fewer risks and more potential for our future plans
I’ll attempt to go through these in a bit more detail for those of you that are interested. I’m also going to start abbreviating because I’m already tired of typing Dell EqualLogic (EQL) and HP LeftHand (LH).
Performance. With the IOPS score as it was I knew that any hardware purchased would need to be able to support this. EQL boxes, unit-for-unit offer better performance and IOPS throughput, it’s as a simple as that. Why did this matter? It mattered because for the solution I was specifying for that meant if a LH solution was implemented an additional box of high-speed (15k SAS) disks would be needed to get the same performance as with EQL boxes running just SATA. The EQL boxes offered sufficiently more performance which meant that on the given budget we could get more capactity with SATA disks in less physical boxes.
This directly affected the next point, price. As a result of not needing extra boxes to increase the IOPS throughput the EQL solution in this instance was about £20,000 cheaper.
The LH boxes are modified Proliant servers – this is good because it means that the server technology is proven. It’s also bad when compared to EQL who have been specifically building storage devices for quite a few years. The LH boxes, in my opinion, would be less stable as “modified” servers when compared to EQLs deliberately designed versions. This opinion was also backed up by various industry experts.
The final point is a bit softer in terms of impact, and what it boiled down to is this: with the EQL boxes we’d have an easier time doing things like box-to-box replication from other sites and therefore simplify our future DR plans and backups.
In order to be fair to LH there were genuinely some things that they did better, but they didn’t affect the business requirements that I was specifying for. They were:
- Bundled with Virtual Storage Appliances (VSAs) – these are little bits of software that allow you to easily turn your old servers with disks in into storage arrays just like the “real” ones you’ve just bought. This provides an excellent way to do cheap backup / disaster recovery
- Network RAID – or asynchronous replication between partners. What this basically means is if you’ve got two boxes usually the data is spread across both boxes, one box fails, you lose everything. With Network RAID you can have the same data on both boxes (which also means it takes up twice the amount of space) – so for super-critical data you can mitigate against the failure of a single box.
All of this was summed up by one vendor’s advice which was:
If you want performance go with Dell EqualLogic and if you want network raid go with HP LeftHand
Did you only consider Equillogic and Lefthand? Did some of the lower end NetApp arrays get considered (i.e. FAS2020/FAS2040)?
I get what your saying about controller performance and Equallogic and Lefthand both having quasi-all-in-one solutions, but NetApp’s architecture is just as scalable. The backend disks can be used with any controller. To upgrade a controller to a higher-end model, it’s just a matter of trading in one controller for another – no data migration.
The price point on the entry-level NetApp arrays should be there as well. The FAS2040 is killer in terms of price and performance. You should really check these out.
Moreover, the space savings and performance gains that can be had through NetApp’s deduplication technology is definately worth taking a look at.
… IMHO
For this particular specification I did only consider EQL and LH, NetApp were a consideration much earlier on but soon got forgotten.
However, in the interests of getting a full picture I’ll actually go back to NetApp and find out what would have been a suitably-sized solution, their software licensing model is quite a large disadvantage in my opinion as it increases the complexity of the final solution.
NetApp aren’t that fussed about talking to me now I’ve made up my mind, which is understandable, but a shame.
As I say though, from what I’ve seen so far the NetApp deployment would be more complex: the software licensing is complicated and scaled whereas in EQL/LH you get everything at once regardless. Also the traditional architecture means that although in theory it would be easy to scale up I imagine that in terms of physically connecting new controllers and arrays it would be a nightmare, not to mention difficult to establish when you needed a new controller, no?
I wrote a similar post covering EQL v NetApp here: http://verydodgy.com/blog/2009/03/why-i-chose-equallogic-over-netapp.html
The lower end NetApp boxes have too many limitations/caviats for small/medium businesses.