Pillar Data Systems – breath of fresh air or a waste of time?

Not knowing much about Pillar Data Systems as a company or their technology I was caught by their 6 minute pitch during a keynote presentation at Storage Expo titled “Standing on the shoulders of giants: the future of storage technology”. They talked about how they’d built their product “from the ground up” as a result of listening to the needs of real storage users, rather than having a 10 or 20 year old product line that is based on an old architecture that has been upgraded and patched countless times in an attempt to meet the demands of the modern storage world.

This idea of building a subsystem from scratch, designed for todays storage needs, got my attention. Based on their speech, their product range promised to be a storage professionals dream come true, the answer to all of our customers needs. I couldn’t wait to see it! Show me one! Give me one! …… Unfortunately on arrival to their stand they didn’t have a techie free to talk to me right away, and as I had other seminars to attend I decided to settle for some glossy marketing materials that would serve as light reading on for the train journey home.

As it turns out they only have one product (box), the Pillar Axiom 500, a midsized modular storage array. Not sure how a single box can solve all of my storage needs (I have so many), still I thought Id read on. As it turns out they appear to have spent a good deal of their time tackling ILM needs. The two main things the glossy brochure was making a song and dance about were -

  1. offering SAN and NAS in the same box
  2. offering different tiers of storage on the same disks (rather than having one tier on FC and another on SATA etc)

SAN and NAS in the same box is not exactly ground breaking, and come to think of it neither is their approach to tiered storage on a single disk. The basics of their tiered storage on a single disk approach is to position data (I assume LUNs) on a disk based on the performance requirements of the data. So data with fast access requirements is written to the outer most tracks of a disk where I/O performance is best, and gradually getting closer to the centre of the disk as speed of access requirements become less important. They call this “tiered storage at the disk level”. Essentially short stroking a disk without the wasted space towards the centre of the disk. Access to the LUNs placed nearer the centre of the disk is also subject to algorithms that give it lower priority to the LUNs on the outer tracks. All of this is supposedly managed through and easy to use policy engine and GUI. OK simple.

Despite none of this being earth shattering, and not looking like its going to solve all of my storage needs, I must confess to still being quite interested. However, as always, a box is only as good as the software to manage it (vendors take note!). As Ive never actually worked with any Pillar kit I cant comment on the software, but gone are the days when Im “sold” on a piece of kit based on its speed or coolness - good software to manage it is now a must!!

Anyway, moving on I noticed that Ben Rockwood, on his great blog, is condemning Pillar to the scrap heap of annoying startup companies – its an interesting read.  Ben gets annoyed that the marketing people at Pillar are using the word “revolutionary” when describing their box.  Although I do side with Ben on this one, Im a little less eager to “cast stones” at a company based on what their marketing people write.  At the end of the day the marketing people are always going to write like that.  However, the main point I would like to make in response to Bens post is this - do Pillars ideas have to be new, or even their own to be revolutionary?

Citrix didn’t invent the idea of the dumb terminal and ESX certainly cant be credited with the idea of virtualising servers and abstracting the underlying hardware. However, both companies have built successful businesses around these concepts and are both arguably revolutionising data centres around the globe. Now Im not saying that Pillar are about to revolutionise the storage world, in fact Im reasonably sure they wont. However, lets not knock them just because they didn’t think of all of their own ideas and their marketing folks like to use words like “revolutionise” and “unique”. After all, most of the big IT companies are guilty at one time or another of nicking ideas from somebody or somewhere else. You can hardly talk to a mainframe guy without them telling you how many of the popular technologies in the open systems world have been lifted directly from the mainframe world.

Another thing that Ben didn’t like was the “used car salesman” approach their sales people seem to be taking, always undercutting the opposition. And again I have to side with Ben, up to a point, on this – none of us like that approach. However, on that note, another thing that I quite liked from the glossy brochure was the idea of a single software license (no hidden costs further down the line). To me that’s a refreshing idea coming from a storage vendor. To see the other option read snigs great post on the greed of software licensing.

So to sum it up I will be keeping an eye on Pillar and looking for opportunities to get my hands on their kit. Please anyone with any experience or opinion with Pillar kit speak up – Ben said in his blog he is open to be wrong, and as for me Im still interested.


Just quickly… One thing that I did find slightly fresh, albeit rather unimportant, was their approach to naming components within their subsystems. If you’re not a fan of acronyms and cant tell your HDDs and DDMs from your DA’s and ACPs then you might like the Pillar approach to this. The three basic components of their subsystem are the controller, called the “slammer”, the policy engine called the “Pilot”, and the drive enclosures called “bricks”. Not as important as how well the kit works but made me smile nonetheless.