EMC Corporation acquired Isilon in November for $2.25 billion. Unlike EMC, Isilon is a major fixture in the HPC industry. The deal closed December 22, 2010 and the supercomputing community may be wondering if Isilon will continue to be an HPC leader.
insideHPC: Will Isilon continue to deliver HPC solutions and be involved in the industry?
Nick Kirsch: Without a doubt. EMC bought us to be incremental where are strengths are. We intend to fully participate in HPC events. And I would say even more so, since we have additional investment and resources. We’re going to move even deeper into those segments where we’re successful today.
insideHPC: What are the additional resources and how will that impact your HPC business?
Nick Kirsch: We have greater marketing resources, like taking advantage of a much larger global sales team to get in front of accounts we couldn’t reach previously. We can accelerate our own R&D efforts. We’ll also pursue additional performance in the product and additional client side access options such as InfiniBand on the front end or PNFS from a performance protocol access standpoint. Also, we’ll grow considerably. We have about 550 employees today. The target is 1,000 employees by the end of 2011, doubling the size of the organization in less than 12 months. These new hires will be targeted at sales and R&D.
insideHPC: What are your HPC plans for 2011?
Nick Kirsch: We’ll have a big presence at SC’11. We’ll be at NAB and BioIT World in April. High performance applications have come out of these vertically targeted shows. We also focus on government agencies through shows like FOSE and DoDIIS. We’ll have new products in 2011 that will improve performance for HPC customers. With additional resources in the technical marketing organization, we’ll address workflow analysis for HPC.
insideHPC: What gap did EMC hope to fill by acquiring Isilon? EMC already has a popular NAS solution with Celera. How does Isilon differ?
Nick Kirsch: Celera is very similar to a NetApp 7G. It is not a scale out model. It’s a great scale up architecture that’s deployed for an enterprise data center for applications and workflows that are data service intensive. Customers use Celera for back office IT, Exchange and SQL. The other part of their portfolio that they’re filling out is that Isilon has done a good job building up its lines of business operations. We understand the non data center customer. We understand the HPC, life sciences and media workflows and these are not areas where EMC has traditionally focused. Additionally, EMC sees an opportunity with what they refer to as “big data.” Over time even general IT organizations are going to have to deal with a massive amount of data.
insideHPC: Who do you see as your competitors?
Nick Kirsch: In HPC, it’s NetApp Data ONTAP 7G, Panasas ActiveStor Modular Cluster System and the Do-it-yourself Lustre or Gluster open source options.
insideHPC: Why should HPC customers use scale-out NAS from Isilon rather than a Linux white box storage cluster?
Nick Kirsch: Our HPC customers are probably more adequately described as commercial HPC. I don’t think we play into the peak HPC market where all they care about are speeds and feeds. We play where the customers want it fast but they also care about it being easy to use and scaling effectively without having a lot of complexity.
Forrester compared the cost of managing an Isilon system versus traditional storage. In HPC, Isilon typically competes against a Lustre or Gluster solution, where an administrator has to be aware of a bunch of different islands of storage and glue them together. The Forrester study found that in that sort of model an administrator can manage about 100 terabytes per person. For an Isilon system the same administrator can manage a petabyte. So Isilon gives its customers a 10x increase in management simplicity.
A lot of the traditional HPC storage systems are fairly fragile. They often rely on a lot of disparate, complex parts. So they have a system that doesn’t stay online all the time. It’s kind of a science project. You have customers who want that absolute performance and are willing to tolerate down time. The customers that appreciate Isilon are the ones that have no tolerance for down time.
insideHPC: Life sciences and media make up the lion’s share of your sales. Why will HPC remain important to Isilon?
Nick Kirsch: Thirty three percent of our revenue is from the media and entertainment industry, 18 percent is life sciences. Many of our HPC customers are in the M&E and life sciences markets, as well as more traditional HPC installations. If you look at our 1600 customers, depending on how you want to specify what HPC is, I believe more than 100 of them are taking Linux clusters of a certain size mixed with Linux environments of a certain size to push through massive amounts of performance. HPC for us is a horizontal that cuts across our verticals.
insideHPC: How is Isilon responding to the needs of HPC customers?
Nick Kirsch: Our customers who are using us for HPC workflows are stressing our systems and they’re doing operations where they want to see us evolve in the most radical ways. For example, we have HPC customers who hit us with large scale compute clusters and they want to be able to move compute operations to our system. Our traditional enterprise customer isn’t saying I want to run compute loads on my storage system, but our HPC customer is. At the bleeding edge of HPC, those customers running us against the massive clusters pushed us to take Isilon beyond 96 nodes to 144 in a single file system. They need more than 3 petabytes and we take it up to 10 petabytes. And now our life sciences customers are asking for 100 petabyte systems. And all that time we want to keep it as easy to manage as possible. We don’t want to turn it into a science project.