Apples and Oranges?

By | April 25, 2007

The apparently unanswerable question was asked, yet again, this time on a rupturedmonkey forum – EMC DMX 3 versus HDS USP  

Im not sick of hearing the question, but I am sick of not getting any “good” answers.  The answers that are out there are either from Hu or Chuck, and to be honest, as good as these guys are, they don't live in the trenches and I think we all know what to expect from each of them.  Then there's the odd magazine article out there that also tend to theorise about architectural differences such as cross-bar switches versus point-to-point networks and almost invariably end up sitting on the fence saying that its like trying to compare apples with oranges. 

But how can it be like comparing apples and oranges?  They are both Enterprise storage systems, both with FC adapters, cache, disks, disk-to-disk copy features, distance replication, built in resiliency…… the list goes on, and one rarely has something that the other doesn’t.  

What I'd love is to hear from people with solid hands on real world experience with both pieces of kit.  In real world day to day operations, does one stand out, or are the both equally as good or bad as each other? 

I specialise in Hitachi storage and my knowledge of EMC is limited to sangod and a couple of customers who have moved from EMC to Hitachi.  These customers are very pleased, but considering they have migrated away from aging EMC kit to brand spanking new Hitachi kit, I wouldn’t expect anything less.  Not exactly a fair comparison. 

But for those with experience, Id like to know –  

  • Is one architecture more or less flexible than the other?
  • Is one noticeably more reliable than the other?
  • Is the software from one vendor way better than the other?  (or are both equally as crap)
  • ……

Of course I don’t expect we’ll ever bury this one once and for all –  I remember a while ago my wife liking her car more than mine.  Hers was more comfortable than mine but mine was so much faster with a much better engine.  But this didn’t matter to her, she liked her fancier dashboard and extra space and turned a blind eye to the crappy engine.  Her car outshone mine in some respects and mine outshone hers in others.  Is this the case with the USP and the DMX?

Nigel

8 thoughts on “Apples and Oranges?

  1. Storagezilla

    What’s usually missing from the “Which is the best?” type of question are the words “..for the problem I’m trying to solve”.

    The point/counter point game of architectures can be kicked around until the sun explodes. It means nothing if it’s not framed in the context of addressing specific issues and it starts all over again with every rev of hardware.

    Both EMC and HDS have material taking shots at each other’s architectures (I’ve read both) but ultimately customers buy for one of two reasons. It addresses their problems and/or it was cheap.

    When buying gear myself I prefer the former over the later. Buying something just because it was cheap is a financial issue and usually completely unrelated to solving a technical problem.

  2. Jesse (SanGod)

    When I was with EMC there was a saying. “No one ever got fired for putting EMC in the datacenter.”

    I’m not so sure if that’s true anymore, our CIO was not a happy-camper when he found out how much the total cost of connecting a single host to the SAN, (when you include PowerPath license, per-port costs, HBA’s, cables – that’s excluding the cost of the raw storage)

    He was even more upset when he realized how the previous CIO had married us to the architecture pretty much irrevocably.

    That’s the catch. Once you start down a path, it’s *VERY* hard to chance course.. And kind of like steering a ship, the longer you spend going the wrong way, the more effort it takes to get you back on course.

  3. stephen2615

    I have a story to tell about storage. It might seem quite bizarre but when you think about it, it served a purpose. I am not quite sure when EMC bought out 2 Gpbs capability but I would take a guess it was about the time HDS did. This happened before I started…

    A project needed outstanding performance and the vendor had just started selling smallish “workgroup” arrays that had 2 Gpbs ports (four off) and 36 GB 15k RPM disks. I think it was the first to be on the market and it promised very good performance. So, as the project needed 20 TB of storage, 20 of these arrays each with a JBOD were brought as the suggestion of the vendor. That’s about 1.6 million in hardware. No one ever thought of anything else as the hardware was quick and no one else could achieve the speed. The 9980V or XP 1024 are all 1 Gpbs internal and it was dismissed because it had no speed… or thats what the vendor said even though they sold the HDS product. No one thought of reliability….

    The system was actually very quick but shockingly unreliable. Each failure caused long outages. I put the problem down to Quick I/O not handling a controller failover. When that happened, 8 TB of storage was unusable due to the complex striping over the arrays.

    When we needed to triple the capacity, the same vendor wanted to sell use another 40 of those systems. When I put my foot down and demanded a USP 1100 or a fully blown XP 12000, it caused a rift between the vendor and my employer that I was quite prepared to ignore. 40 systems that would cost 3.2 million. I don’t think so. I got a USP 1100 with new Cisco director class switches will lots of change left over. The vendor was not happy but hey, tell someone who cares.

    I managed to physically migrate the old storage to the new storage in one day and never looked back. What did we do with the old storage? It was virtualised and made a good backup staging area. No Quick I/O installed on those systems and we did not care about controller failover or disk failures any more.

    So, the lesson was ignore any sales droid even if you have known them for years and like them. Always go out to the market when looking at huge dollars and attempt to screw your old vendor for everything that are willing to do to keep you as a client.

    At my new job, I personally don’t think we get a lot of use out of our USP’s as we have Windows systems that can barely push a feather into the breeze. Our dev and pre prod systems run on relatively cheap IBM DS series hardware and they run well. I believe we only have USP’s as they are pretty well bullet proof and provide excellent reliability which is what we have to have to ensure important information is always available.

    So, if the DMX 3 (or even 4 which could be on the cards) was considerably cheaper when it come to tender, I would not hesitate in buying it as it is very similar architecturally and it probably stays up just as well. When your company wants to get a foot in the door, amazing deals can be done.

    End of ranting and raving..

    Stephen

  4. nigel

    stragezilla – I hear what your saying. However, I feel that many customers often base their decision on the latter rather than the former of your points. This is probably because so many customers see the two products as being so similar that cost can often be the only easily distinguishable factor. This is especially when the sales people get involved, as they are not as impartial and honest as Chris has been in his response to my post over at (http://storagearchitect.blogspot.com/2007/04/whats-your-favorite-fruit-emc-versus.html).

    I think Stephens reply above is evidence that customers see little if any real difference in the products, may be except from the virtualisation aspect. In fact I know of HDS customers who have historically been big EMC shops who tell me that they invited HDS in to beat up on EMC who were starting to take it for granted that they would buy their storage from EMC – to shake things up a little and keep vendors on their toes so to speak.

    Chris – no probs about hijacking my post. To be honest its exactly the type of thing I was looking for.

    Jesse – Ive also heard the saying that nobody got sacked for putting EMC into the data centre 😉 On a similar note I also know of people who’ve looked at bids and felt that those from the big iron vendors was “reassuringly expensive”. Often some of the not so big players can come in with lower bids, offering similar solutions, but the lower cost can put people off sometimes. You get what you pay for.

  5. TimC

    The *please enter authentication code* but strikes again. *sigh* I love when technology *just doesn’t work*.

    Anyways, I’ll be far less eloquent and try to retype this.

    I think the biggest problem you’re going to run into is the industry itself. What are the odds someone is going to go out there, learn the ins and outs of a storage array, learn every last niche of it, and then find a new entry level position starting on a completely new piece of hardware. If I am an EMC expert, I’m going to sell myself as such. How many people out there are looking to hire a *storage admin* who has never actually admin’d the storage they are using??

    Now, you may have the shop who moves from one to the other, but how many shops make that migration before the technology is old? You don’t drop 5million on storage and then change your mind 6 months later.

    So really, I think about the best case scenario you’re going to find are people in test environments. IE: QLogic or Emulex or Cisco or Brocade. They are the most likely people to be unbiased, and have both arrays in a lab somewhere testing away. BUT, testing != the real world, so even then… good luck.

    And fix that damn authentication code!!! 🙂

  6. stephen2615

    I was thinking about this virtualisation thing. We already have a fair bit of storage virtualised behind the USP and HDS are really really pushing to put all our other storage behind it. The pro’s and con’s are pretty clear cut but I never thought about HDS making it very hard to move to another platform. If we did decide to go EMC in the future, then what about the virtualised storage. It would be a nightmare to revert it to non virtualised storage or try to do something along the lines of EMC & Cisco virtualisation strategy. I for one don’t want to get involved in that. So, unless EMC do it for free, they are not in the running to get into our storage market.

    I am unbiased and I don’t like Brocade switches at all. Give me a MDS (or even a McData) switch any day of the week……

    Speaking of storage engineer jobs, no one cares what you have. They are just sooooo happy to get someone. When someone with only a few months playing around with Storage Navigator and knows how to zone a couple of hosts in a Brocade switch can get $150k, what can I say. I think it is criminal when you have to take over the mess that that moron left.

    But that’s another story.

  7. Jesse

    You know – The thing I hate to hear more than anything is “Brand-A is faster than Brand-B when you set them up in identical configurations”

    Well of course they are. Different architectures are built to different requirements. EMC and Hitachi both have their “sweet spot” as it were.

    The true test of one vendor over another is if you set each of them up in their “optimal” configuration. Max out the cache and stripe in whatever configuration it was designed to work best using. *THEN* give me a raw performance number and I’ll judge who is “best.”

Leave a Reply

Your email address will not be published. Required fields are marked *


*

You can add images to your comment by clicking here.