Isilon Chief Speaks Out

It's been a tough 12 months for Sujal Patel. Just six months back, the founder of Isilon landed back in the CEO's seat (not his first choice) when his newly public company failed to thrive, prompting the board to reshuffle its executive deck and re-evaluate various aspects of its sales organization.

A new CFO and a self-imposed financial audit subsequently turned up accounting irregularities. Last week, Isilon revised some past earnings downward by about $7 million and issued its latest financial report in an effort to clear the air and start fresh.

It's a transition for Patel, who views last week as the closure required for Isilon to move on. We caught up with him last Friday, after a grueling week in the Wall Street hot seat. Here's what he told us:

Byte and Switch: How is it going? How do you see things shaping up for Isilon now?

Patel: I think of last nights call as a great transition for Isilon, closing one chapter of our history and getting through a difficult period for all our stakeholders in 2007. I’m absolutely excited about the market opportunity that I think Isilon has in front of it, totally excited about our position competitively in the marketplace.

Figure 1: Sujal Patel, CEO, Isilon Systems

Byte and Switch: I'd like to ask you a couple more questions related to resolution of the problems. How did this get by you?

Patel: I don’t want to dwell on the past too much here and in the form 10-K we filed with the SEC a few days ago, we spent a lot of thought on the disclosures that went into that... I’d refer you and readers to that document for details, but essentially it was weakness in our revenue recognition.

Byte and Switch: I read the document, but I'm interested in how that happened. Some say Isilon went public too soon and was unprepared for some of the rigors of the reporting required. Could you speak to that?

Patel: I certainly wouldn’t say that we went public too soon. I think that the company's leadership position in the clustered storage space as well as the momentum that we’re seeing in the space are really quite significant. I think the challenges of 2007 can be attributed to many things and with respect to the actual revenue recognition issues, we describe those in the disclosures in the 10-K. I’d refer you back to there, unfortunately.Byte and Switch: OK. So where do you go from here? Describe what’s on your roadmap for the short term.

Patel: From here I’m pleased to say that I get to focus my full attention on building a business and building this company into a leader in the clustered storage space over the course of the coming years. What I view for us through the year here, as I mentioned on the [April 3 earnings call], is a huge opportunity for us to invest back into our products and to really enhance them in a number of key areas.

One opportunity is really enhancing the core platform that we have, and one of the announcements we made in January was that we released the Isilon X series, which is our next-generation storage platform, capable of scaling to 1.6 Pbtyes in a single file system for storage capacity and up to 10 Gbytes/second of throughput. [See Isilon Steps on the Gas.] That is a significant number more than traditional NAS and SAN systems. Through the year, you’ll continue to see us make updates to our storage system platform as well as to our One-FS core file system and operating system software. And then as well over the course of the next four to six quarters, you’ll also see us continue to add features and functionality to our core products as well as to our software application suite.

Byte and Switch: You mentioned Isilon's core file system. Could you speak to some of the issues? Seems like file systems are taking lot of mindshare these days.

Patel: You know, the core issue that customers are facing out there is that traditional NAS and SAN storage systems just don’t allow you to centralize a tremendous amount of capacity into a single file system, and they’re inflexible in that they can’t give you the ability to scale the capacity of your file system and the performance of your file system independently.If you think back to the reason Isilon was founded over seven years ago, we were founded to take advantage of a trend that was beginning to take hold in the data center: That trend was clustered computing, the ability to take multiple industry-standard servers and combine them together into a single system leveraging high-speed networking.

As that trend really took hold in the data center over the course of the last few years, it’s become apparent that on the storage side there needs to be an architectural shift that can keep up with the capacity and the performance of these new clustered environments, as well as these new virtualized compute environments. And that’s really what clustered storage is all about -- it’s a technology which is based on file system technology that enables you to combine multiple systems, what we call nodes in our architecture, into a single cluster, so that you can create a highly scaleable storage system both in capacity and throughput. And while you’re creating that system, you can add automation functionality to make the job of administering a large storage system -- hundreds of terabytes in a single network drive -- a simple process.

To Page 2

Byte and Switch: So where is Isilon now regarding addressing? Are you ahead of the curve in terms of what customers need? Are you behind it?

Patel: I think that we’re in a great place right now, and I feel like the market for clustered storage is coming at us so quickly that we will continue to invest to capture [it].Let me give you some interesting statistics: We’ve been selling our clustered storage systems for roughly five years now, and the thing that’s really exciting is over the course of those five years we have nearly 750 customers and have shipped over 10,000 clustered storage systems. The interesting thing that we’re seeing today, and something that I mentioned on our conference call yesterday, is that we’re seeing a marked increase in the number of larger enterprises that are looking to adopt and are adopting clustered storage systems from us.

That’s really interesting for us, because the wider NAS space is a few-billion-dollar market today and growing; the wider SAN space is two to three times that size and growing today; and both of those spaces are really dominated by large enterprises -- at least domestically, they are. Internationally, of course, the statistics are similar, but... As we see mainstream enterprise customers adopt clustered storage, we see a huge market opportunity in front of us, and what we’re doing is investing in our products, in our service and support capabilities, and investing in our business, so that we’ll be able to support large enterprise customers. And we’ve done a successful job over the latter half of 2007 positioning the company to be able to handle that penetration into the enterprise.

Byte and Switch: Why do people need all this clustered storage? Why do they need to be able to scale performance and capacity separately?

Patel: Let's dissect this in a few different ways. At a macro level, the most important trend to think about here is something that Gartner Group recently described in their research, and that is that unstructured data or content is growing faster today in enterprises than structured data is. That means we’re seeing a huge shift in terms of the type of data that enterprises are struggling with today, and that shift is only going to continue over the course of time.

The other thing that’s interesting out there is that all the major enterprises are going through a transformation in terms of their data center architectures. Data center architectures are moving toward clustered computing. You’ve got lots and lots of physical servers which are very similar to each other, and they spread applications out among those servers.If you think about the success of VMware and the virtualized computing environments that are out there, this is one of the applications that leverage clustered compute power. So virtualized computing is a major, major trend and one of the trends that’s driving clustered computing in enterprise data centers.

The thing that happens, though, when you have a huge amount of compute power and you have lots of physical servers is you need to be able to feed those servers with a large amount of capacity and as well significant performance. What that means is that clustered compute requires clustered storage to be able to keep up with the needs of the data center. That’s one of the major reasons why we’re seeing such need from larger enterprises today.

In terms of applications, we’re seeing a wide range of applications that are pushing the envelope in terms of capacity and throughput. Within the oil and gas exploration space, of course there are folks out there -- major oil companies and their customers --who are crunching large data sets to look for new oil deposits, new oil fields, analyzing seismic information that’s coming in because oil is such a expensive commodity and profitable business. In the manufacturing space, we see lots of momentum because of the huge increases in CAD/CAM environments and computer-aided engineering in creating everything from airplanes and airplane engines to cars. In bio-IT and the life sciences space, we’re seeing a tremendous amount of neurological research, cancer research, drug discovery applications, all of which are driving the performance needs on the compute side of things. And we continue to see very strong momentum in our core vertical market that we penetrated five years ago, media and entertainment -- film production, broadcast television, and so forth.

Byte and Switch: Is clustered technology the only way to approach this? Other vendors are doing different things. Why is your way better?

Patel: One of the interesting trends that has been true now for many decades is that industry-standard hardware, commodity hardware, can be coupled together to create very powerful systems that always come up with the best price/performance ratio and always the best architecture to bet on.If you think back to the early '90s, this was the exact trend that Network Appliance took advantage of to become a successful company initially. Back in the early '90s, a company named Auspex was the undisputed leader of the network attached storage space. They used a lot of proprietary hardware. They had a system that was much more difficult to innovate because of that. Network Appliance came along and based their systems on industry-standard hardware. They were able to innovate rapidly and grow the business rapidly, and of course they are a leader in the NAS space today.

We view that same sort of trend existing today. Clustered computing as a platform and high-speed networking that backs that up is clearly the right architecture for data centers going forward, and all of the analysts we talk to see that today. For example, Gartner mentions in that same survey I mentioned earlier that by 2012, they expect that NAS systems with clustered file systems will account for 40 percent of the market.

If you look at existing major vendors like Network Appliance and EMC, they’re both validating clustered storage as an architecture. I think that everyone sees that this is the direction in which you need to go. But the real key advantages that we have as a company that position us in an incredibly good place are that one, we have had the opportunity to be working on clustered storage and putting significant efforts and investment into our innovation over the course of the last seven years.

The second thing as well is that we had the opportunity to go and take a blank sheet of paper and reinvent the architecture for clustered storage, so we’re not held down by the legacy of traditional monolithic storage products that are in the portfolios of companies like NetApp, EMC, IBM, Hitachi... all the other major companies. So what we are able to bring to customers is a really unique value proposition and one that really resonates well.

Byte and Switch: Do you worry about NetApp and EMC? In all financial analyst notes, they talk about Isilon facing strong competition from those companies, as well as from DataDirect Networks and some private companies.Patel: Yes, of course, being good businesspeople, we are always paranoid of our competitors. Network Appliance and EMC are absolutely the two competitors that we focus on; they are the vast majority of the field engagements that we run into.

The exciting place that we are in today is that both Network Appliance and EMC are talking about clustered storage, they are validating it out in the marketplace, and they’re creating a lot of momentum. But today, if you look at those vendors, their product offerings are in a very, very early stage. EMC has yet to release a clustered storage offering; they’ve been talking about a project [codenamed] Hulk and Maui for quite some time now. On the Network Appliance side, they have a product, the OnTap GX operating system for their core filers, which is still in a very early stage and isn’t really seeing enterprise penetration because it’s really not ready for that yet. The vast majority of our field engagements with Network appliance are still on their 7G product line, which is their traditional filers that are not clustered.

To Page 3

Byte and Switch: What about DataDirect Networks? Are they really presenting a competitive problem?

Patel: In actual field engagements, the vast majority of the time, we’re competing with Network Appliance and EMC, and those are the two guys we’re focused on. I think that when you look at some of the other vendors that are private vendors, the thing that you don’t see and the reason we don’t run into them is that as we continue to penetrate the enterprise and try to move clustered storage into this mainstream adoption phase, companies that are private and smaller aren't able to penetrate into those sort of accounts, and really as we continue to push clustered storage’s ascension into that mainstream data center, I think that we’ll continue to see the large competitors as opposed to the smaller ones.Byte and Switch: You were talking about enterprise requirements, but you see the bulk of what you address as being large content, unstructured, and media streaming. Is that right?

Patel: I wouldn’t use those words to describe it. What I would say is that unstructured data is actually a very wide type of data. It encompasses things like digital content, which is video, audio, images, graphics, but it also encompasses scientific data, it encompasses CAD/CAM data; it encompasses a wide range of data types that we see out there.

To give you an example, we have customers in the e-discovery space who are storing millions and millions of PDFs and Word documents and documents that they’re analyzing and indexing and stretching and accessing at various times. It’s those sort of applications that people don’t traditionally think of as unstructured data, but they really are unstructured data. And because of the fact that unstructured data environments are growing faster than structured data, we view all of those as a huge opportunity for us.

Byte and Switch: Would you ever focus on structured databases?

Patel: I think that clustered storage architectures will ultimately be used for all types of data in the storage market. IBM recently bought a company which is focused on clustered storage for block data, for transactional data like databases. It’s not a focus area for us and it won’t be in the near term, but certainly in the long term if I look out in our five-year vision, certainly you’ll start to see us move in that direction as well, but certainly not in the short and mid term. We view the opportunity that we have in storing unstructured data and storing file-based data as so large that we really need to focus on that opportunity.Byte and Switch: OK. I wanted to ask you about storage networking. Do you see any particular trends: 10 Gbit/s Ethernet and iSCSI versus and Fibre Channel and Infiniband, and so forth?

Patel: Within enterprise data centers, we are certainly seeing that 10 Gbit/s Ethernet and Ethernet-based technologies are beginning to rise as the winner both on the block interconnect side, i.e., iSCSI, as well as on the file-based side using protocols like NFS and CIFS.

Byte and Switch: Do you see an increase in InfiniBand at all, given that 10 Gbit/s is still expensive?

Patel: We certainly see InfiniBand as an important technology to connect compute clusters together, and to that degree we’re definitely seeing momentum in InfiniBand within enterprises.

I don’t know if you remember or not, but actually the backend interconnect of our own clusters is InfiniBand and has been since April 2005. It’s been a huge success for us. We’re definitely seeing momentum on the InfiniBand side of things as well, and it’s usually on a compute cluster side.Byte and Switch: Do you think eventually that 10 Gbit/s iSCSI will win out ultimately in the data center?

Patel: Ultimately, I believe that is the case. The Ethernet-winning trend has been going on for decades at this point. Token ring, ATM... Ethernet has always seemed to have won out in the end. We don’t see any difference there with respect to Ethernet capabilities on the block services side of things.

Byte and Switch: To what extent do you think social networks and cloud computing will affect what you do?

Patel: Two different questions there. Social networking is a very interesting phenomenon that is going on right now, and it’s driving huge amounts of storage requirements within the Web 2.0 space and the Internet space, and frankly, that’s just great for vendors like us, because we provide a lot of those Web 2.0 companies the unique capability to centralize their data and grow it in a very easy manner, which helps them to have an unrestricted business model.

On the other side of things, cloud computing is beginning to get a lot of play out in the marketplace, and it’s an interesting technology; it’s an interesting solution that I think will begin to see some momentum out in the marketplace with particular applications, where it’s applicable and where it works well. Cloud computing, of course, is predicated on the fact that you have to access data over a network, so to the degree those networks are fast enough for particular applications, I think cloud computing will be able to make some inroads.One of the things that’s exciting from my perspective is that cloud computing is going to require storage behind it, and we have the opportunity to be a major part of the storage solution within service providers who are providing cloud compute and cloud storage capabilities.

Byte and Switch: Do you think that over time service providers will be a larger component of your customer base than today?

Patel: I think that it’s definitely possible. One of our target core markets is carriers, which we view somewhat broadly as telcos, cable providers, wireless, and definitely we’re seeing sort of utility storage service providers emerge both domestically and internationally. We have some of those customers especially internationally, where I think this trend occurred first, and there is some overlap with the carrier segment that we focus on. But certainly I think that in a more general way, the service providers that are providing these cloud capabilities certainly could be an interesting target for us and one that we think about.

Byte and Switch: Anything else?

Patel: No... I am really excited about the opportunity that Isilon has. I’m excited about seeing the clustered storage space that we have been talking about for so long really begin to become a mainstream technology.Have a comment on this story? Please click "Discuss" below. If you'd like to contact Byte and Switch's editors directly, send us a message.