BlueArc Hopes Its New Server Will Bypass the Web Bottleneck
By DON CLARK
Staff Reporter of THE WALL STREET JOURNAL
MOUNTAIN VIEW, Calif. -- Much of the history of the high-tech industry could be summarized as the search for the next "hot box." Enrico Pesatori thinks he has found it.
The well-traveled computer executive left Compaq Computer Corp. last year to lead a start-up called BlueArc Corp. that is generating an unusual amount of buzz in Silicon Valley -- and proving that economic gloom hasn't slowed the race to attempt to change the world with powerful computer hardware.
Like a new face in Hollywood, BlueArc has intrigued some star-making analysts and investment bankers, despite the fact that it has only shipped a few sample copies of its unusual machine.
"They've got breakthrough technology," says Robert Gray, an analyst with International Data Corp. "Other people will have to go this way."
George Gilder, a technology prognosticator with a fondness for cosmic metaphors, writes in one of his newsletters that the BlueArc box "reconciles the time domain of the Web with the outer space of the fibersphere."
But other silicon-based starlets are waiting in the wings. And all the hopefuls must audition against some technology troupers, the big companies in a long-hot market for computer servers and data-storage devices that shows signs of cooling.
Computer "box" stories usually start with a problem. BlueArc is focusing on a bottleneck in the flow of Web pages, video clips and other files around corporations and the Internet.
Where slow networks once limited the transfer of that information, many of those data highways are now much faster -- so fast that the current generation of systems that store and retrieve files can't keep up. The situation is a bit like ditching your Ferrari at the freeway onramp to ride home on a bike.
Geoff Barrall, a British engineer who founded the company as Synaxia Networks in 1998, attacked the problem by departing from the basic architecture used to build computers since 1945. Designed as general-purpose calculating engines, those systems rely on a central processor that carries out a series of tasks determined by software.
But things evolved differently in communications networks, where designers of traffic-switching boxes such as routers have long reached for ultrahigh speed by using chips that are tuned for a single purpose. Mr. Barrall, who used to design such networking devices, believed that specialty chips also could handle repetitive tasks in data storage and retrieval, a main function of servers. These special chips do such things as interpreting the basic technical protocols and filing schemes used by popular operating systems. So instead of employing a Jack-of-all-trades microprocessor as its brain, each BlueArc server uses an assembly line of specialized chips that do a few tasks at blinding speed.
In benchmark tests at its headquarters here, a rack full of BlueArc components fetches data files at five to 10 times the speed of competing products. And the company claims two other dramatic consequences: About 100 times more users can tap into one server than other storage appliances, and each server can manage up to 200 trillion bytes of data, or 30 times more than rival machines.
"This is absolutely disruptive technology," Mr. Pesatori proclaims.
Not everybody agrees. Some say BlueArc hasn't identified the right problem, let alone solved it. Customers are just beginning to validate the box's performance on real-world tasks. And BlueArc's machines still need to be attached to computer servers that run software such as databases and do the actual processing of information. That could introduce delays.
Regardless of its impact, the company's quest underscores some major changes in the computing landscape. For years, from Apple Computer Inc.'s first machine to Seymour Cray's initial supercomputer, innovative general-purpose computers were the ticket to stock-market success.
PCs changed everything. Intel Corp. began churning out millions of microprocessors for those devices, and makers of large and small machines started using those commodity chips rather than developing their own proprietary circuitry. As manufacturers standardized on other key components -- including programs like Microsoft's Corp.'s Windows -- computer makers with original designs became an endangered species.
But some specialized box makers did build markets. EMC Corp. of Hopkinton, Mass., introduced storage systems that packaged processing power with arrays of disk drives. In the 1990s, Network Appliance Inc. of Sunnyvale, Calif., helped popularize the concept of appliancelike storage servers that can be shared by multiple networks, not just serve up files to one system. That's much more convenient for companies with complex networks, and the device came along just as the Web caused storage demands to skyrocket. Economic worries have hurt both companies' stocks lately, but EMC's market capitalization still stands at more than $70 billion, while Network Appliance's is greater than $7 billion.
So venture capital has flowed to storage-related hardware. Established companies and start-ups are now racing to deliver the next hot box in the field.
To name just a few: Tricord Systems Inc., based in a Minneapolis suburb, is trying to slash costs by clustering many storage appliances with software that manages them like a single data repository. Nishan Systems Inc., Fremont, Calif., develops storage switching devices that are based on Internet protocols, which are easier to deploy than the standard ways data is usually pumped to storage systems. A secretive start-up called Zambeel Inc., also in Fremont, this year plans to deliver storage hardware providing "virtually unbounded scalability and performance," vows marketing director Greg Marek.
But BlueArc may stir the most controversy. A presentation by the company at a Lehman Brothers investor conference in San Francisco Thursday attracted a standing-room-only crowd. Mr. Barrall gamely answered a string of questions, but some audience members seemed to have trouble understanding the new box and where it would fit in the crowded market.
One questioner raised a pivotal issue: Other makers of storage servers use Intel chips for their underlying intelligence, betting that the giant chipmaker's huge research budget and economies of scale will bring speed improvements at the lowest cost. While some of the data pipelines between those chips and disk drives are relatively slow, major improvements should start arriving this year.
BlueArc's so-called Silicon Server design is a contrarian bet -- that conventional-style storage devices will never adequately close the performance gap. These systems inevitably run into delays, Mr. Barrall argues, as their microprocessors execute software and retrieve data from memory chips to complete storage tasks.
The company's boxes are expected to go on sale in the second quarter at less than $100,000, a price designed to compete directly with Network Appliance. Shridar Subramanian, a marketing director at that competitor, argues that reliability and other issues worry customers more than performance. He also questions whether special-purpose hardware can be updated quickly enough, and whether an ultrafast storage server would find itself slowed down by the other servers on networks that run business applications.
"At the end of the day, the bottleneck may shift from one place to the other," Mr. Subramanian says.
BlueArc designed its machines using chips that can be quickly reprogrammed by customers to fix bugs or make other changes. And, while the company is initially focusing on the storage market, it may eventually adapt the same principles for an ultrafast machine dedicated to running a single program, such as an Oracle Corp. database.
One obvious target, BlueArc says, will be companies that hope to broadcast audio and video over the Internet. But Appshop Inc., a Fremont-based computer-services company testing the new box, is particularly excited about the speed with which the system could handle the time-consuming chore of "backing up," or creating a copy of, huge databases, says Mike Jennings, its vice president of operations.
Altera Corp., which is supplying chips for the machine as well as testing it for its own use, also sings its praises. "My systems administrators came back and the comment out of their mouths was 'Wow, this thing is fast,' " says Frank Hannig, chief information officer at the San Jose, Calif., company.
Mr. Barrall, a 32-year-old Ph.D. who originally focused on robotics, initially caught the eye of Ottawa-based Celtic House International, the first of three venture firms that have invested $37 million in BlueArc. It recruited Chairman Gianluca Rattazzi, an Italian-born entrepreneur who started three Silicon Valley companies and helped woo Mr. Pesatori, who holds the chief executive title and had worked at Olivetti SpA, Zenith Data Systems, Tandem Computers Inc., Digital Equipment Corp. and Compaq. The 130-employee company started in England and still conducts most manufacturing there.
"As soon as we understood the concept, we said, 'That is a very big idea,' " says Andrew Waitman, Celtic House's managing general partner.
Write to Don Clark at don.clark@wsj.com |
