back to article The shoemaker, the array refresh and the VMworld smackdown

If the shoe fits then wear it; it’s the motto you'd expect a footwear retailer to do business by. Designer Shoe Warehouse (DSW), a 400-outlet US shoe chain, applied that advice when its main storage array, a VMAX we are told, failed to deliver data fast enough and began costing too much. It bought an InfiniBox instead and the …

  1. Nate Amsden Silver badge

    seems like a bad config

    I mean an array that can only deliver 33k iops at 300ms? Must've been doing something very wrong. Never used VMAX myself but that number seems just bad.

    Also 177k iops from cache on infinidat? Also seems bad, would expect at least 3 to 4 times that. Should be able to get more than that with just ssds pretty easily.

    Maybe these folks have some really strange workloads but those numbers make little sense to me.

    Without more details this is not a case study I would trust.

    1. Bronek Kozicki Silver badge

      Re: seems like a bad config

      Note, this is a firm with 11,000 employees - I suppose there is some serious load on the filer there. Quite possibly multiple databases servers running at the same time on the same store, too. Definitely lots of VMs (they said as much).

      1. P. Lee Silver badge

        Re: seems like a bad config

        11k employees, 2k VMs = 1vm per 5.5 employees

        I must be old. Back in my day, we 'ad 'undreds o' workers per server down t'mine - an' we were glad of it!

  2. Mystic Megabyte Silver badge
    Coat

    The shoemaker, the array refresh

    I read that as the shoe fetish. Mines the PVC mac.

  3. liquidh20

    Total marketing rubbish. Clearly a bad VMAX configuration. A XtremIO or VMAX-F would kill these figures. Even an Unity-F can do better figures than both combined.

    1. Bronek Kozicki Silver badge

      It might be bad configuration, agreed. The trouble is, their service told them repeatedly that the configuration is fine, then they upgraded FC from 4Gb to 16Gb and replaced VMAX kit as well, and the performance was still bad. Which tells me that someone selling this kit made a good money, yet the client was still suffering from poor performance. Which means that yeah, if the configuration was wrong, the both the sales and service were totally shit and do not deserve to be ever used again. Hence, move to a different platform makes sense - even if all their problems might or might not have been solved with a well advised configuration tweak instead.

      1. Anonymous Coward
        Anonymous Coward

        EMC needs a competitor that's hungry, VMAX is a terrible product IME

        EMC is slow, lies about their performance, and the VMAX is vulnerable to all kinds of Bad Things because they refuse to update it to support things like, I dunno, SHA2, high levels of encryption on the HTTPS services. Oh, and are you HIPAA compliant? not if you setup the EMC management portal as they have DESIGNED it. Basically, they can (and do) get in at any time to your storage management, and more than once EMC has updated things without permission or warning - service affecting things.

        But do, by all means, continue to waste your money on their "white glove" VMAX experience, people. Just be aware that that white glove is made of latex.....

  4. Anonymous Coward
    Anonymous Coward

    Its still just a Cache box...again

    Infinidat is Moshe's 3rd box... this was a catfight between his 1st (Symmetrix) and 3rd (Infinidat) designs. However the common theme in all of Moshe's designs is a dependency on DRAM cache. They shove up to 3 TB of it in these 3 Dell servers for nothing but write cache. Read cache is handled via SSDs in the storage controllers. However the backend storage is ALL slow 7.2K drives. The reason why infinidat looks good versus AFAs is because all of the AFAs stopped using dram for cache and just use their SSDs or NVRAM. So yes it is going to be faster on some tests, as long as the cache has room. They also switched from a dual controller design to a triple active controller design held together with an infiniband mesh. More CPU and More memory with every controller having access to all of the drives. Its a combination of the much larger write cache and triple controller design that is giving Infinidat the leg up right now. However if you exceed what the cache can handle you quickly enter territory where an AFA would provide a much more predictable performance workload.

    My only worry with infinidat is that like all of Moshe's companies is that he is going to prop this up only to quickly turn around and sell it off. Then run on to start yet another company and take his talented staff with him again. XIV has stagnated as a result of this very move. Then you have Dell right now as their OEM provider for the storage controllers. With the EMC merger complete I can't see them playing nice for encroaching on their largest revenue storage product line. So the infinidat array is going to probably end up being re-architected with another vendor in the near future. (HPE?)

    1. Rob Isrob

      Re: Its still just a Cache box...again

      "re-architect with HPE?"

      huh?

      They state they are COTS. You understand what that allows? This for example:

      https://storagebuilders.intel.com/membership/infinidat

      "New technologies, such as Intel® 3D XPoint™ and Supermicro NVMe* connectivity solutions, provide new advancements and opportunities. "

      They are already kicking or about to kick Dell out the door. Wouldn't anyone else in the storage space that had a lick of sense and embedded OEMed Dell be looking to Supermicro or another low-cost tier1 whitebox vendor?

  5. Anonymous Coward
    Anonymous Coward

    It's all about the services

    You buy a VMAX for the data services, an XtremIO for speed, a VMAX AFA for both. Buying an array from a startup is unlikely to work out very well in the long term.

    1. Throatwarbler Mangrove Silver badge
      Thumb Down

      Re: It's all about the services

      You buy EMC for "reassuringly expensive" kit with a long pedigree. If you want a good ratio of price:performance, you go elsewhere.

  6. g4n

    Not a fair fight!

    Take any enterprise storage system from 3 years ago, and you can find a dozen better alternatives in 2016, even from EMC. Come on!

    Today, for VMware loads you can even go without a storage, get any performance you like, and have stretched metropolitan high availability as a welcome bonus (i.e. all flash VSAN).

    But 3 years ago? In comparison, we where in the stone age!

    I don't love spinning rust, but you are not allowed to spit on it today, just because we have a viable flash option: it' not fair, and almost disrespectful.

    Just plan a quick transition to anything current, thank the old system for the high availability (not so much for the performance), and learn yet another lesson useful for the next big wave.

    Just my 2 eurocents...

POST COMMENT House rules

Not a member of The Register? Create a new account here.

  • Enter your comment

  • Add an icon

Anonymous cowards cannot choose their icon

Biting the hand that feeds IT © 1998–2019