DataDirect Networks Finds Room at the Top
For DataDirect Networks, petascale computing is nothing new. They’ve been delivering petabyte-size storage systems for years. Last year the company shipped 30 petabytes of storage and that amount will probably more than double in 2007.
DataDirect Networks, or DDN, was founded by Caltech graduates Alex Bouzari and Paul Bloch in 1998 with the goal of providing high performance and high capacity storage for the most demanding computing applications. The company’s timing was perfect. In the late 90s, the digital data explosion was taking off. From 1999 to 2002, stored information was growing at about 30 percent per year. That trend has actually accelerated. The latest results from IDC’s Worldwide Disk Storage Systems Tracker showed that 1,000 petabytes worth of capacity was shipped for the second consecutive quarter of 2007, an increase of 60 percent over last year’s figures. The growth of multimedia on the Internet is clearly fueling a lot of this. But all applications on desktops, servers and HPC systems are mining larger and larger data sets.
Disk technology is racing to keep up. While CPU performance is being driven by a Moore’s Law doubling of transistor counts every 18 months, disk capacity is actually doubling about every 12 months. But according to Josh Goldstein, DDN’s VP of Product Marketing, that doesn’t necessarily give storage the upper hand. In fact, while disk capacity, processor performance and network bandwidth are all increasing at rapid rates, disk drive performance is stagnant.
“The rotational speed of the drives hasn’t changed much in ten years and seek times have not changed at all,” explains Goldstein. “You’re basically dealing with the laws of physics with regards to how fast you can move the mechanisms inside the drives.”
The result is that external storage is increasingly becoming the bottleneck to computation. Finding a way to design around the mechanical limitations of disk drives has been the ongoing challenge for all storage hardware vendors. And the technical challenges are especially severe for capacity- and performance-driven applications — the types of codes commonly found on high performance clusters, supercomputers, and ultra-large-scale, Google-like computers. These are the kinds of storage environments that DataDirect lives for.
The company’s claim to fame is their Silicon Storage Appliance (S2A) technology, which is common to all DDN’s storage controllers. S2A is based on a real-time state machine, where data is processed in parallel through FPGAs. Since the controller logic is implemented in hardware, rather than CPU-bound software, DDN is able to collapse the storage infrastructure, integrating switches, virtualization, RAID, caches, and disk-level optimizations into a single box. One controller can have up to 960 disk drives behind it. With one terabyte drives on the horizon, this means that a single storage controller will soon be able to manage nearly a petabyte of data.
The FPGA implementation also helps bring enterprise-class data integrity to SATA drives, which are especially popular where capacity cost is the driving factor. Unfortunately, SATA hardware has data integrity limitations since it doesn’t implement parity checking natively, making it prone to silent data corruption — not a good thing if you’re, say, performing a nuclear weapon simulation. S2A technology takes up the slack by performing the parity checking for the SATA drive. If data corruption is detected, the controller repairs the error on the fly. Since this is all done in real time with hardware, no performance hit is incurred.
If the drive becomes “sick,” the S2A controller will take the drive offline and attempt to restore it to a good state. If successful, the only write operations that need to be redone are the ones that were journaled while it was offline, avoiding an entire drive rebuild. The whole idea is to guarantee data integrity on SATA drives so that they act more like expensive fibre channel drives.
By making lots of cheap storage look like lots of expensive storage, DataDirect has carved a comfortable niche at the high end of the market. In the Top500 list of supercomputers, 7 of the top 10 machines (including the number one Blue Gene/L at LLNL) and 35 of the top 100 are all on DDN storage. Less storied systems are supported too. Merrill Lynch (financial modeling), Synopsys (EDA tools), Ford Motor Company (CAE) and Shutterfly (photo sharing) all use DDN storage, while AOL and Fedex are using DataDirect gear for high performance backup storage systems.
The success of the company in the high performance computing market has drawn the attention of several HPC vendors. IBM, Dell, Cray, SGI and Bull all use DDN storage for some of the systems they sell. With last month’s announcement by IBM, the company’s Deep Computing group will now sell DDN gear under a rebranded DCS9550 name.
The relationship between IBM and DDN was inevitable. ”For the longest time, we had always competed neck-to-neck with IBM in the HPC storage market,” said Alex Sayyah, DDN’s Director of Marketing. “Rather than constantly having to waste resources competing with each other, both our companies saw more synergies and benefits in a mutual collaboration. That’s how that relationship started.”
The hope is that DataDirect can expand into other IBM accounts where the benefits of the S2A technology make sense, such as commercial HPC, or where HPC-like storage performance is required, such as large-scale disk archiving.
To keep up with the customers’ ever-expanding storage needs, not only is the company selling more systems, but the amount of storage per system is increasing as well. Last year they booked $86 million in revenue; this year they expect to reach $100 million. Since 2002, DDN has enjoyed 19 consecutive quarters of profitability, and in the past 4 years they’ve exceeded a 52 percent growth margin. If that sounds like a company that’s ripe for a public offering, you’re right. Ernst and Young is in the process of auditing the company’s financials to get them prepared for an IPO.
While many storage vendors are struggling to stay profitable, DataDirect Networks seems to have found the right combination of market focus and technological differentiation to become successful. At a time when storage is becoming the center of attention for many high-end users, DDN finds itself in the spotlight. For them, storage has become a true growth industry.