Posted by: Peter Burrows on September 13, 2007
Keep an eye on Xsigo Systems, a new start-up that just had its coming out party after three years of developing its product. It makes what it calls an I/O Director, a piece of hardware that it hopes will become a staple in the world’s data centers—which represent one of the fastest growing markets in tech.
On the face of it, that’s a tough challenge. Big boys such as IBM, HP, Sun, Cisco and EMC are already battling hard for this market, and aren’t likely to let a newcomer waltz in. But Xsigo isn’t taking on any of these giants directly. Rather than compete in servers, storage or networks, its product is focused on a different task—interconnecting those three key legs of the data center stool.
That may sound mundane, but it’s not. In recent years, companies have been able to boost the utilization rates within each of those domains. Brocade and others have helped improve storage utilization through storage networking. And VMWare is the hottest new IPO, because it helps companies virtualize their servers so they’re left doing nothing far less often. But what about when a certain job requires more storage througput, and less network bandwidth? Say a bank wants to go from handling transactions—a server-intensive task—during the day to running analytics—a more storage-intensive task—at night. First of all, you’d have to add a slew of $1500 interface cards to make those connections possible. And someone would have to make like an old-fashioned Ma Bell operator and start changing cables around every night. “As technology advances, the bottlneck moves to another place,” says Xsigo backer Vinod Khosla, a co-founder of Sun who has gone on to a legendary career as a venture capitalist. “Now, the bottleneck has moved to I/0 (or interconnect).”
Khosla says one of the things he likes about Xisgo (pronounced See-Go) is that it’s got a sales pitch that works on many levels. At first, he thinks the company will land business by simply saving CIOs money. He claims operating costs (not the cost of new technology) is 70% of most CIO’s budgets. As such, negating the need for all that manual cable-pulling addresses the biggest cost center.
But should it succeed in getting in the door, Xsigo also has a larger strategic story to tell. At a time when data center operators are trying to move to utility computing (or some variant of that imprecise phrase), Xsigo’s technology can be a key enabler. Utility computing requires that a data center be able to adapt on the fly to whatever jobs need doing. It’s not enough to just have have all the servers behave like one virtualized pool, and the same for the storage and network. The holy grail is for the entire facility must move as one—as though it were one big mainframe, rather than a collection of various products from various suppliers. Xsigo’s products aim not only to simplify the connections within this virtual mainframe, but add the software intelligence to manage the changes.
Khosla says this ability to appeal to CIOs on differnet levels is a hallmark of the successful tech firms he’s been involved with. Sun, for example, got traction by simply offering a cheaper workstation than rivals; only later did larger “the network is the computer” strategy become apparent.
No doubt, there will be competition. In particular, there’s Cisco, which spun-in a company started by former engineering chief Mario Mazzola. Evidently, it’s also working on ways to replace the many types of connections that now crowd the data center (eg. FiberChannel, Infiniband, Gigabit Ethernet) with one unified fabric.