Already a Bloomberg.com user?
Sign in with the same account.
For the first time in years, hardware startups are trying to break into the market. Their gambit: Inexpensive special-purpose machines
Steve Dewitt remembers an incident that occurred soon after he joined startup Azul Systems three years ago. He walked out to the driveway of his Silicon Valley home to pick up the morning newspaper and ran into a neighbor -- Bob Evans, a legendary former IBM (IBM) executive who oversaw the development of the mainframe computer in the 1960s that led to Big Blue's industry dominance. Dewitt had recently told Evans of Azul's plan to create a radically new server, and Evans had been enthused.
But now that Evans had a few days to mull the plan, Dewitt recalls Evans telling him: "You know, Steve, I hope you realize how incredibly hard it is to create a new computer company. There's just so much that can go wrong."
That's why for the hundreds of tech startups that have come and gone in the past few decades, only a small percentage have tried to build new computer companies. True, plenty of networking startups were created during the buildout of the Internet. And some have even made important technical advances in storage and the packaging of computers. Traditional servers have been shrunk down to the form of so-called blades, about the size of a pizza box and nothing more than circuit cards with powerful processors that can be slipped side-by-side into racks.
LESS CARE AND FEEDING.
But when it comes to fundamentally rethinking the basic gear that makes up the typical corporate data center, not much has happened. Indeed, only one new hardware company has truly hit the big time since 1990: storage systems maker Network Appliance (NTAP).
Yet for the first time in years, a number of ambitious startups are dead set on beating the odds. They're betting that a new generation of powerful chips, the ability to focus on one task -- say, storing data or Web traffic -- and an agnostic approach to operating systems can combine to provide big cost saving. Many of these newfangled devices are built with low-cost, off-the-shelf parts, such as Advanced Micro Devices' (AMD) Opteron chips and garden-variety disk drives, just like industry-standard gear from Dell (DELL), Hewlett-Packard (HPQ), and others.
But these machines also feature innovative software, designed to help them accomplish a particular job with far less care and feeding from info-tech staffers. That's critical, since 70% of many companies' IT budgets go into maintaining old gear rather than adding new capabilities. "Our industry has trained its customers to put up with so much pain" in terms of complexity, says Steve Goldman, CEO of Isilon, a startup in this new generation. "That makes it easier for us to look good. Like the old saying goes, 'if you want to look tall and thin, hang around with short fat people.'"
At Azul, Dewitt has turned the blueprint he showed Evans into a server built around a powerful new breed of chips known as multicore processors. Azul's machines are designed to run any software developed with Java-like virtual machine technology -- without a care for the age-old question of whether the server is based on Unix, Windows, or Linux.
Azul just started selling its new machines and is drawing interest from a wide range of customers desperate for lower-cost, more efficient ways of running their IT shops. "General-purpose servers have gotten faster, but just incrementally so," says Steve Lapekas, chief information officer for Pegasus Solutions, a Dallas-based processor of hotel-bookings information. Thanks to the Internet, Pegasus must now handle 16 online inquiries for every room that's booked, up from a four to one ratio a decade ago.
To try to manage all this traffic -- none of which brings in any additional revenue -- he's taking a chance on Azul. He plans to use two of its servers in January. "With these, I can replace almost 10 racks holding 400 servers. It could be very disruptive."
WORKING OUT KINKS.
With tech budgets barely growing but Net traffic continuing to skyrocket, other hardware startups are also finding customers willing to give them a shot. Much of the action centers on storage, which will be the biggest spending priority for companies in 2006, according to a new CIO survey by Sanford Bernstein. Startups including 3PAR Data and Isilon are making fast progress with customers such as MySpace, Kodak (EK), and NBC (GE), which is using Isilon gear to archive video during the Olympics in Torino, Italy. Customers such as Amazon.com (AMZN) and Cingular are using data-warehousing appliances from Netezza that let them do detailed data-mining of consumer buying trends.
Of course, these small fry are no great threat to the kings of the data center just yet. Many of them are still working out kinks with their technologies, and all are building sales and support organizations to keep big customers interested. As Ann M. Livermore, head of HP's $31 billion corporate computing unit notes: "Is Citicorp going to take a chance on running its business with one of these companies?"
Indeed, HP has its own big plans to help customers do more with less, by using its growing suite of management software to remake the data center of the future. "We want to reinvent the economics of IT," by automating many jobs and simplifying the job of running data centers, says Dick Lampman, the head of HP Labs.
IBM is also seeing rapid growth for its blade offerings. It has been forging partnerships related to its BladeCenter technology, so that many aspects of a traditional data center can be included inside a single rack -- not just IBM servers but also Cisco (CSCO) networking gear, Brocade (BRCD) storage-networking switches, and even servers running Sun's (SUNW) Solaris operating system. IBM also continues to build its own automation software (see BW, 12/5/05, "Computer Heal Thyself"). Plus, analysts expect a warm welcome for its recently introduced z9 mainframe (BW Online, 7/27/05, "IBM's Mainframe with a Mission").
And Sun is hardly standing still, either (see BW Online, 12/7/05, "Sun's Super Servers"). Last year, it brought out new servers built around AMD's Opteron chips that offer improved energy efficiency for some computing jobs. On Dec. 6 Sun followed up with systems built around its new T1 chip, which puts eight separate processor cores on one piece of silicon. And to increase the number of programs built to run on these new hardware platforms, Sun has also decided to give away much of its software, from Solaris itself to a host of middleware.
Still, the general-purpose server is coming under attack from a range of more specialized gear. Call it the parade of Network Appliance wannabes. Back in the early 1990s, that company figured a way to lop off one job -- storing files on networked devices called filers, rather than keeping them trapped inside disk drives on a particular server. By separating the storage of data, companies create "network-attach storage" (NAS) -- a more efficient approach than having servers that were always either painfully empty or overstuffed. And by using off-the-shelf components to keep costs low, NetApp was able to maintain 60%-plus gross margins -- providing a blueprint for other hardware entrepreneurs.
BEYOND BIG IRON.
Scott Weiss, the CEO of Ironport, sought out NetApp CEO Dan Warmenhoven's guidance in creating his maker of e-mail security appliances. And he thinks many other entrepreneurs will follow suit. "The general-purpose server is akin to the electric motor 100 years ago, which Sears used to sell with 20 or so attachments," says Weiss. "After a while, companies came along to make far superior fans, toasters, and dispose-alls."
This new approach is taking off mostly with so-called Web 2.0 companies, which are dealing with unprecedented challenges as they try to develop full-blown applications via the Web, rather than just static Web pages. Consider the plight of Aber Whitcomb, CIO of MySpace, the red-hot social-networking site frequented by indie rock fans and legions of teens. MySpace is currently adding a million new customers every six days -- and it has no idea when that growth will slow.
That means Whitcomb can't afford to keep buying big iron every few years, as was common in the past. Instead, he uses new, clustered storage gear from 3PAR Data to process traffic and is starting to use equipment from Isilon to store certain kinds of content. One reason: Both let him add modules as he goes, with minimal human interaction. "We're always on the lookout for new hardware approaches because we have to be on the cutting edge," says Whitcomb. He figures he would have had to buy 40% more conventional storage to match the capability he gets with 3PAR.
"THE LINES ARE BLURRING."
Other startups have even bigger plans: To alter the basic layout of the data center as it's known today. Rather than focus on making one component of a server faster -- say, a speedier processor or a bigger storage bank -- these companies are focusing on eliminating bottlenecks within these various kinds of gear. Fabric7 Systems in Mountain View, Calif., just began shipping a new kind of server that's also loaded with high-speed networking technology and software.
That way, when a customer receives an online transaction, the Fabric7 server can not only process the order but can also set up the appropriate Web connection, whether its an e-mail, or a higher-bandwidth link to connect a voice-over-Internet-protocol phone call, or even a supersecure link to complete a billion-dollar trade for a Wall Street bank. "Servers, switches, networks -- all the lines are blurring," says Sharad Mehrotra, who developed servers for Sun before co-founding router maker Procket in the mid-1990s.
If these hardware pioneers succeed, they could make things painful for today's traditional server companies. After all, one of the newcomers' goals is to help companies get more out of less hardware. "In reality, the $20 billion storage market should be a $10 billion industry," says David Scott, CEO of 3PAR Data, whose storage gear has so-called thin provisioning software that lets companies use only as much capacity as they're programs actually demand rather than set aside vast amounts of extra room just in case traffic spikes.
For a possible glance at what's ahead, consider Savvis, even though it's struggling mightily to deal with some big problems. It's burdened by losses related to the U.S. assets of Britain's Cable & Wireless, including 15 big data centers. And it's coping with the scandal surrounding its former CEO, Robert A. McCormick, who left Savvis after allegedly running up a $200,000 tab at a New York strip club.
But before he left, McCormick oversaw the development of a "virtualized service utility" that's built with storage gear from 3Par and servers from upstart Egenera. This infrastructure lets Savvis continually adjust how its servers, storage banks, and network capacity is allocated, according the needs of customers such as Reuters and Proctor & Gamble (PG). Rather than run their own data centers, customers can order what they need -- less on holidays, more when they're closing the quarterly books.
Indeed, while it normally takes 90 days for a company to prepare its data center to roll out a new application, Savvis claims it can do that in minutes. "We can take an electronic order...and have it up and running in five minutes," says Vice- President for Product Management Vincent Dimemmo.
BACK TO BUSINESS.
Will a new crop of industry powers emerge? Many will undoubtedly be purchased, and many will fail. But a number of these Hardware 2.0 pioneers say they've noticed a pickup in interest in the last 90 days.
Azul's Dewitt thinks CIOs have finally finished making sure their companies have the IT systems to comply with financial-disclosure regulations, such as Sarbanes-Oxley. "They've spent the last two years in compliance lockdown," says Dewitt. "But I've sat down with 15 or 20 CIOs in the last month, and it's clear they're getting back to the business of trying to stick it to their competitors -- which at the end of the day, is what you're supposed to do with information technology."
Dewitt admits that while potential customers like Azul's vision, they're still waiting for proof that his company can deliver. "Look, we have an installed base of none, so we have no right to be arrogant." But if his technology -- and that of other computer entrepreneurs -- comes close to having the impact he predicts, the industry could be in for exciting, tumultuous years ahead.