Exclusive Interview of Alex Bouzari, CEO of DataDirect Networks

DataDirect NetWorks in Facts and Figures

Company
DataDirect NetWorks, Inc. (DDN) formerly MegaDrive Systems, Inc

Headquarters and offices
Chatsworth, CA; Colorado Springs, CO; Columbia, MD; Mountain View; France; Germany;  Japan; United Arab Emirates

Date founded
1988 for MegaDrive, 1998 for DDN

Financial funding
$12 million in 2001, then financial funding shares acquired between 2004 and 2007

Revenues and profitability
$180 million in 2010, $127 million in 2009 for fiscal years ending December 31
Operating income: 13% of revenues in 2009

Other main executives

• Paul Bloch, president, originated from France with French and U.S. citizenship (Bouzari was born in Paris but is U.S. citizen)
• At the board there is also Jean-Luc Chatelain, VP and CTO of the Enterprise Information Management practice at HP, from the same country. Consequently conversation at board’s meeting are frequently in French.

Number of employees
320

Average system price range
Around $200,000

Technology partners
Fujitsu,  Hitachi, Toshiba, Seagate

OEMs
Bull, Cray, Dell, HP since six months, IBM (6620 for SONAS), SGI, Sony

Number of active customers
500

Number of systems
Around 1,500 shipped this year

Average capacity per customer
250TB to 500TB, increasing 100% per year

Main customers
For OEMs Dell and IBM; for end users Lawrence Livermore National Lab (more than 100PB) and OakRidge National Lab

Applications
HPC (57% of the Top100 HPCs), then media (social networking, broadcast, post-production, etc), life sciences (genomes, research in hospitals), security

Competitors
EMC/Isilon, IBM, LSI, NetApp, Panasas, SGI

Alex Bouzari

At 49, he is CEO of DataDirect NetWorks. He brings more than
twenty years of experience in the storage industry. Prior to co-founding
the company with Paul Bloch, he served as CEO of Personal Writer,
and co-founder and president of MegaDrive Systems (also with
Bloch). Bouzari holds Bachelor of Science degrees in Engineering and
Economics from the California Institute of Technology (Caltech), with
graduate studies in Engineering at MIT and Stanford University.

Q&As

StorageNewsLetter: Like DDN president Paul Bloch, you came from MegaDrive Systems in the same activity sector. What happened to that company?
Alex Bouzari: So MegaDrive was started as a systems integration business. And we ended up taking the customer base that we had developed as MegaDrive and shifting it into DataDirect. So MegaDrive morphed into DataDirect. With different products, different technologies.
What’s the percentage of revenues for HPC only?
HPC is 40%.
Why do you concentrate on HPC, a relatively small market within the IT sector?
Because we could differentiate. So HPC has very different data patterns from the broader market. We could develop specific technology to serve those types of requirements. And then HPC led to rich media because of sustained bandwidth requirements. And then when the Internet happened, it just became a much bigger market. We went from a very small market, being a niche player, where we were number one, to a much bigger market where we now have, it’s a $4 billion market, and we have 5% of it, the market became much bigger. Initially it was great for us because we had very little competition, because it was very specialized, very difficult to do, very difficult to earn the trust of customers.
Why haven’t you had the same success as Isilon, given that you offer the same kind of solutions for broadcast and entertainment?
Well the main reason is from the beginning, we wanted to have a profitable company. We were profitable in $25 million in revenue, and we ran it very profitably all along. We were always focused on profitability as opposed to growth.
After HPC, you got into imaging?
So social networking…
And surveillance?
Surveillance also, we have some very large projects in surveillance. City surveillance, subways, airports, borders.
What’s your secret technologically that enables you to attain one million IO/s and up to 10GB/s throughput from a single storage system?
It’s a highly parallel architecture, lots of pipes going to the server, lots of pipes going to the disks, and we aggregate performance of the disks and serve them to the server. And there’s a lot of artificial intelligence algorithms to do pattern recognition on the data flow. So when data comes in, we can parse the data, separate metadata from data, if there are bottlenecks in certain pipes we can redirect the traffic in real time. There are a lot of techniques that we’ve developed over the past few years.
Do you use standard RAID, unlike Isilon?
It’s a modified RAID-6 architecture for real-time applications. Because most of our customers are real-time.
I suppose you use a lot of RAM.
Not a lot. We have lots of processors inside the appliance to do the number crunching. So it’s not cache-intensive, it’s processor-intensive.
Is it patented?
Yeah, we have a dozen or so patented that are either granted or underway.
For your Web Object Scaler Object-Based Cloud Storage System with global name space, do you use your own software?
Yes. We spent four years in development, started shipping this year.
Is it any relation to Quantum’s Stornext?
It’s completely different. This is to handle a very large number of objects in a distributed fashion, so billions of objects with real-time replication to handle internet-type applications.
Do you have de-dupe?
Yes.
Your own technology?
Web Object Store technology. Yes, it’s our own algorithm.
What kind of ratio?
It depends on the application. We deal with images and in many cases, our customers aren’t truly interested in losing some of the information that come with the image. So it’s not something that really applies to our market. But if you look three years from now, when the market expands, it will be useful then.
On xSTREAMScaler 4.0 with Quantum StorNext, you have de-dupe. With your own algorithms?
Quantum StorNext is their technology. We have third-party
integrated solutions, where we use the capabilities of the third party,
and over the last four years, we started developing our own software. So
it’s starting to become a little complicated because you have
third-party technologies, and our own technology.
What’s the future for HPC: IB or Ethernet?
Still very much IB. So if you look at 2014-15, it’s still very much IB for the very large systems, mainly because of cost and scale. So 10GbE is still much more expensive than IB.
That can’t be…
For large, large networks it’s still much more expensive. So everyone who is doing very large-scale HPC deployment is doing IB. And IB architecture for large-scale HPC applications is still more than 70% of what people do.
Why would you say that IB is used primarily in HPCs?
Because HPC customers are very cost-sensitive. And all they care about is performance and scale. More cost-sensitive than enterprise customers.
There are some advantages to Ethernet, though, wouldn’t you agree?
Well, one of the issues with HPC is latency. The fatter the pipes, the better it is in terms of latency. So every few years, they try to do 10GbE, but the cost is never right. And the problem is that people like Cisco don’t want to lower their price because it affects their other markets, and they don’t want to make an exception for HPC.
SAS or SATA HDDs or SSDs? What do you sell, percentage-wise?
Today it is SAS for performance and SATA for capacity. I think that by the end of 2014, it will be solid state for performance and SATA still for capacity. But solid state will cross over SAS in about 4 years. It’s a small percentage, in price it’s less than 10%… We work with everyone, Intel, etc. It’s an industry that is still evolving.
It’s not well known that you are also in VTL. How many have you sold?
We probably do, something in the 10s of petabytes, but it’s for specific applications where customers are looking at a fully-integrated solution. It’s not replication  –  archiving applications, those types.
Your roadmap?
It varies by market. For the image-based environments, social network, media, it’s more software capabilities…
Management software?
Management of the network in a distributed way. So these types of customers typically have end users who are distributed worldwide, and so the management of the content across locations can be very complicated. So it’s anything that can help manage a very large number of objects in a distributed way, and we’re investing a lot in those types of things. It’s intelligent migration from solid state into SATA, in order to optimize the cost and performance. It’s providing integrated solutions between I/Ops and bandwidth. More and more customers want one solution that can handle all of their requirements. So it’s 75% software and management capabilities, and 25% hardware platforms.
Why did your PR division approach me for an interview, given that your firm is not known for great communication?
It’s interesting. At the beginning we wanted to stay under the radar, we wanted to avoid competition from the big guys. Up until $3 million, which was three years ago, nobody knew how big we were. And then we’ve always been more of an engineering company, so more focused on customers and understanding what they need and developing products for them.
Not EMC-style?
Not EMC-style. So whenever we’ve had a dollar to invest, we’ve put it into technology rather than marketing. But now that the company’s growing in size, we’re doing a little bit more. But I don’t think that we are marketing.
Is it because you need broader recognition, in advance of preparing an IPO? That’s a rumor I’ve heard several times these last few years.
We do need more recognition, but not for an IPO. Nothing in 2011, for sure, maybe after, it depends on market conditions.
Did you already try an IPO before?
We looked at it in 2008. But then the markets became very strange, so we never filed. Again, because the company is profitable, cash-flow positive, growing, we don’t have investors pushing for an IPO, it’s really not something that we’re very focused on. Our main goal is continue to grow the business profitably.
And to remain totally independent? Or did you seek a buyer?
This year there were lots of transactions. Many of these people are our customers. I think for us, in terms of next steps, we’re really focused on taking it from $180 million to $500 million. We figure that’s a four-year plan, and we’re focused on executing to that plan. I think that acquisitions are not something that we’re looking at. I think an IPO would make more sense, but only if we think that there are customers we should be buying, that would require a lot of money. Otherwise it doesn’t make sense, it’s a lot of headaches, aggravation, distractions. You have to run it for every quarter, so I’m not sure it’s a good thing for customers either.
Have you acquired any company?
No.
DDN was involved in Tera 10 with Bull for the CEA, but I didn’t see your name for Tera 100.
We did Tera 100, which was three hundred GB a second. We shipped it at the end of 2009, and it was deployed by mid-2010, and it’s operational now, for the past few months. So the CEA has been a good customer for us. We shipped around 50 systems shipped this year.
Is there any business relationship between DDN and the French firm Intellique, where Thierry Bloch, brother of your president Paul, currently works?
They are a support center for us in France. Logistics and support.
A few personal questions. Where were you born?
Paris.
And you are a U.S. citizen?
Yes, so is Bloch. He has dual citizenship also in France.
Hobbies
Art. Classical Music. Opera. Nothing to do with technology. That’s why I come to Europe often.