Analysis

posted on 15 August 2008 11:26

Not that most people knew it was in the clustered storage business but Rackable has decided to sell its RapidScale clustered storage business. Taken with EMC, HP and IBM's slowness in promoting their extreme scalability products it looks as if such products are very hard to get right, leaving Isilon in pole position.

Rackable has had a poor quarter and the development costs of the RapidScale clustered and highly scalable global namespace system are too high compared to the hoped for days of wine and roses. Rackable will partner in future for its Web 2.0/HPC storage products and, since it already partners IBM for blade servers, it wouldn't be too much a stretcher seeing it offering the XIV product.

But there is a problem. XIV doesn't scale. You can currently have up to 80TB of usable space in one rack with no clustering. This is almost ... I don't know what term to use here. IBM's XIV webpage talks of petabyte levels of scalability. As an undershoot, 80TB instead of 1PB is quite impressive. Just why is it that vendors who are onviously spending a lot of money and development effort are finding it so hard to bring products to market and promote them energetically? I'm thinking of EMC (Hulk - InfiniFlex 1000 - Maui software), HP (ExDS10000), and IBM (XIV).

A couple of years ago Isilon burst onto the extreme file storage scene with its clusterable IQ system. That can scale to a nominal 1.6PB of capacity and have 96 nodes in a cluster with a single namespace and very fast access. The system resonated very well with growing file storage demands which became more and more prominent.

There were the famous, even infamous, pair of exploding data universe papers sponsored by EMC out of IDC. There was the rise of Web 2.0 applications and suppliers needing to grow from a few gigabytes to petabytes of capacity, such as MySpace, Flickr, and SalesForce.com. There were movie houses needing massive render farms to create the special effects in movies like the Terminator franchise. There was also the incursion of high performance computing (HPC) into enterprise business IT.

NetApp had its clustered filer stakeholder with ONTAP GX. BlueArc had its supercharged NAS filer silicon. But Isilon's products, better suited to this market, struck a chord and the company grew like topsy, went IPO, and then crashed and nearly burned through poor management, since replaced. Absent that it would now be nudging the half billion dollar a year revenue mark instead of being a recovering $100 million annual revenue company.

But even in its weakened state it is not been given killer competition by anybody. Very recently it announced that four more movie companies were using its kit to make and edit movie files. Is there any Hollywood special effects house that isn't using Isilon? More to the point, why are three huge companies that are spending millions on developing and acquiring similar technology not marketing the heck out of it?

IBM has made its XIV system available but not announced it. Millions were spent buying the XIV business and its 'father of Symmetrix' founder, yet IBM has announced it with all the marketing energy of a feather duster. EMC's Joe Tucci talked in November of last year of Hulk and Maui, hardware and software for Web 2.0 and cloud computing applications, with Maui software likened to Isilon's OneFS software and called a global repository. But nothing has been announced. There were hints of an appearance at EMC World earlier this year and, if you looked hard, there was an InfiniFlex 10000 hardware product, but it was not press announced. The Maui software wasn't present. Since then, silence.

Even now if you search EMC's webpages for InfiniFlex the only hit you get is A 15 Minute Guide to Mainframe Environments, where it occurs in the EMC trademark list. This isn't even feather duster marketing. EMC has discovered the invisibility marketing cloak.

HP made a big marketing splash with its ExDS 9100 Extreme Data Storage system in May. It's going to be available in Q4 this year and HP is happy to tell us about the CPU complex and the storage complex and how any CPU can talk to any storage element but it won't say how. No, HP is not ready to discuss that. Just like EMC is not ready to discuss details of the InfiniFlex. Why this reticence? Why this soft sell?

There could be two sets of problems here. One centres on product development difficulties. The XIV undershoots its marketed capabilities. EMC has not found it easy to develop the Maui software. HP might (I stress the might) be having difficulties with the internal switching in its ExDS9100. The second problem area could be marketing.

Isilon is not winning every deal it goes after and isn't going after every conceivable deal it could. There aren't enough Isilon sales reps on the ground to do that. So many customers, potentially buyers of Isilon/ExDS9100/InfiniFlex/XIV products buy the high-end SAN-type storage arrays instead. They buy an HP XP, an EMC Sym, an IBM DS8000. This gives ammunition to marketeers and product managers who don't want existing revenues lessened by cheaper clustered storage products. If they don't have to eat their own lunch then you can't blame them for not doing so.

If Isilon wasn't so weakened by its internal woes it might have prompted more energetic responses from EMC, HP and IBM. As it is we're left thinking that Isilon has time to recover because three potentially smart and aggressive competitors are pulling their marketing punches. The company has new products coming. Inevitably they will scale more and be faster and give our big three vendors more of a fight. They might rue the day that they gave Isilon time out of the ring to recover because of their own internal product development and/or product marketing problems.

And Rackable? Which storage systems supplier needs to buy a highly scalable clustered storage hardware and software development team and assets? A supplier with no product presumably and that would identify just three possible candidates: HDS; Sun; and possibly Dell, not a good selling environment for Rackable. Making extremely scalable storage work seems to be extremely hard work.

^{[Chris Mellor.]}

Comment:

From Russ: Great story on Rackable and clustering.

HP's product is kind of interesting.... but it's simply repackaged PolyServe around HP gear. Not truly integrated.