Storagebod Rotating Header Image

Isilon – A Wrap?

So it looks like that EMC will get Isilon barring any mishaps and counter-bids; I stand by my reasoning here but I've been thinking a bit more about this and I've had a few conversations with people.

I've chatted to a few vendors and many of them were saying that if they didn't work for their current company, Isilon were their company of choice; they thought it looked pretty cool. Not sure how many of them actually had any real idea about what Isilon were about but they thought it was cool anyway. Obviously once Isilon become an EMC company, their coolness factor will decline. (Sorry Chad, but the videos that the vSpecialists tend to record are bad and not in a good way!)

It's very interesting how quickly the whole scale out storage thing is progressing; everyone wants a piece of the action and Isilon certainly have the scale-out NAS story down pretty well; block, let's be generous and say that it is early days yet. But what was once a scientific and media specialised device is becoming mainstream. Isilon however were looking at ways that they could attack the entry point and how they could attack the Celerra and NetApp filer-space; so anyone who believes that Isilon is a cloud-only play from EMC, well I think that they are wrong. 

In fact, I would go as far to say that in the same way that 3Par will become the foundation of all HP's block storage; Isilon will become the foundation of all of EMC's file-based storage; from Celerra to Atmos, this acquistion will impact them all. Atmos with Isilon becomes a more interesting if not compelling offering than Atmos alone; it is there that I suspect that all of the initial integration focus will be. How do you graft Atmos onto Isilon is probably the next question that EMC's Atmos engineering team have to answer; well once they've worked out how to integrate Atmos and Centera and more importantly how to do so without a mouthful of humble-pie.

Of course this leaves us with some interesting questions about C-NX or whatever it might end up being called; like HP and EVA, EMC have some questions to answer. Still the Clariion is a much under-rated block storage array even by a large chunk of EMC and as Isilon really doesn't do block; I think it's future is safe for some time to come but perhaps it's another integration point. 

I also wonder if EMC will spend time moving the BSD-based Isilon product to their internal common Linux platform; let's hope this does not turn into a NetApp Spinnaker re-tread. I don't think that people should get too hung up on the interconnect between the Isilon nodes; yes it's Infiniband today but there's good reason for that, it was probably the cheapest and most effective low-latency interconnect Isilon could use without spend a fortune on specialised 'R and D'. Isilon's smarts are in software development and not really hardware; EMC could really help push them to the next level of hardware.

EMC's biggest challenge though is going to be taking a company which is culturally very different to EMC (I've dealt with both) and integrating it into EMC without loosing the clever folks that they've just bought. Isilon currently feels like dealing with a start-up but a start-up with mostly smooth corners and I don't think that they need to change too much. EMC might have just bought a company which will have as much impact on their storage offerings as their purchase of Data General all those years ago. 



  1. dlove says:

    i am a current isilon customer with a couple clusters and i’m frankly pissed off at the acquisition. i made the decision to RUN from emc when we were forced to use an acopia/f5/celerra/clariion rats nest half-ass implementation for managing a lot of unstructured data.
    it was a miserable experience overall and we made a decision to move over to isilon at all costs! the f5/arx piece was great, made moving off emc even easier 🙂 we even were able to dump the arx piece which was another acquisition that sucked for a customer (call f5 support and press #9 for “other”, thanks!)
    we now dread the future of this platform and now i regret not going with something “open” like zfs where i’m not locked in with a particular vendor.
    it seems the future is M&As and all smaller/innovative companies are guppies in the sea of storage whales. i thought isilon wouldn’t be a sellout but i guess that’s how it works…
    isilon says “We expect that you will see no change to the leading technology and service you’ve come to expect from Isilon as a result of this transaction. Isilon’s entrepreneurial culture and the rapid product innovation that Isilon is known for will continue.”
    i say, i’ll believe it when i see it…

    1. Tanjidin says:

      Disclosure EMCer here. Bod hope you’re doing well found this blog post fascinating. First of all while EMC pinitooss Isilon as Big Data (most beneficial to customers when their scale or use case demands scale-out), there’s NO DOUBT that scale out architectural models have inherent advantages, and are widely applicable in broad use case scenarioes. In fact I personally would go as far as to agree 100% with your assertion if you are looking to put in place or refresh your file-serving environment in the next couple of years; you are doing yourself and your employer a massive dis-service if you do not look at a Scale-Out solution I do want to make two notes. 1) I’ll pile on the observation that certain classic enterprise NAS features (file-level snapshots, large scale filesystem writeable snapshots, compression, dedupe, etc) are currently not there in Isilon/SONAS and their ilk. These are NOT instrinsic, and I completely agree that adding those features is likely to prove easier than to bolt on than it is to bolt on the core architectural model. Expect a lot on this front.2) With scale-out NAS models, individual IO latency is TODAY generally material higher for transactional NAS workloads. This is based on distributed metadata and data models, and the inherent latency that introduces. Just speaking for EMC, the VNX file stack has much lower small-block IO write latency (for things like VMware on NFS, or Oracle on NFS) than Isilon. Scale out block models (3PAR, VMAX, HDS) use very specific hardware models (generally shared memory) to achieve low IO latency while keeping scale out working. This is the harder area of R&D when it comes to scale-out NAS (as the inherent architectural benefits of scale-out instrinsically make latency trickier the classic your strengths are simultaneously your weaknesses ). This is an area of insane work. If it can be cracked, the possibility of scale-out NAS models to become extremely dominant will be much more pervasive.Stay tuned Thanks again!

  2. Martin G says:

    Really Insane Price?
    Reel In Punters?
    Run Isilon People?
    Rich Isilon People?
    Restful Interface Proposed?

  3. Martin G says:

    Anybody with a innovative product which actually starts to gain traction will be acquired…none of the big boys wants another NetApp. There will not be another NetApp!

  4. dlove says:

    top 12
    ruined its persona
    ripped into pieces
    reduction in personnel
    return if possible
    remedies in place
    research incentive plan
    request in process
    regulatory investigative powers
    ram intake pipe
    request invalid page
    ran into problems
    and finally…
    rusted into pieces!

Leave a Reply

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