According to the running commentary after the announcement, Oracle’s purchase of Xsigo is a step toward SDNs, or maybe it’s to go after VMware in network virtualization, or to boost its cloud credentials. And I thought coverage of the Nicira deal was vacuous! Of course, the Xsigo probably doesn’t do any of these things, and it’s probably not aimed at doing them either.
The root of the problem here, I think, is that Xsigo says it has “virtual I/O”, and these days anything that’s “virtual” is “cloud” and “SDN”. Not so, guys. Virtual storage is valuable in any data center that has a storage pool, and probably the largest market for it is in virtualization-based data centers. And in any case, Xsigo grows its virtual I/O out of its real business—fabric.
Yep, underneath the hype, Xsigo is a fabric switch vendor, one of two (Mellanox being the other) who’s made a name for themselves in a quiet way by exploiting the Infiniband switching technology introduced by Intel. If you looked for the leading technology in data center fabric today, at least if you defined “fabric” in an objective way, Infiniband would be it. Oracle has used Infiniband before, as have other IT giants, and so the move is more evolutionary than radical. But what drives it? I think it’s simple; own the data center.
If you believe in cloud revolution for enterprises, it starts in the data center. If you believe in new network architectures, the data center will be at the heart. If SDN means anything its meaning has to start in the data center, and if you’re a salesperson for network equipment or IT equipment you’d darn straight better be calling on the people who run data centers. The bucks start and stop here. And that’s what Oracle wants, the bucks.
Oracle makes servers, servers they want to make a fixture in data centers. The cloud and virtualization are two drivers of data center technology change and thus two insertion opportunities. Both the cloud and virtualization demand some rethinking of the relationship between storage and servers because a fixed or even traditionally provisioned relationship is too static for fast changes in application-to-resource mapping. That means too slow for the cloud or even for dynamic applications of virtualization. Xsigo offers virtualization of the I/O, so Xsigo solves a problem for Oracle.
So doesn’t this mean Oracle is after Cisco, VMware, and the cloud? No. Going after somebody is helpful only if they have the market you need to get. Cisco is certainly a server competitor but if you want to steal market share steal it from IBM or HP or Dell. VMware doesn’t make servers, so why chase after them? The cloud is in its infancy so if all you can do is sell fabrics to cloud providers you’re in for a long cold winter. No, this is about positioning and issue ownership. What drives change is new issues, because planners want to make sure that their long-lived capital investments actually DO live long. Oracle is grabbing two hot things; fabric and virtual I/O. They’ll follow the trail of the cloud to the bank, but they’d follow other trails too. The cloud is only a means to an end. Same with collision with Cisco or VMware; if they’re in the way, you collide, but they’re just collateral damage.
Rounding out tech, we had our first Google+ hangout yesterday, a format we’re calling “Techout Live” ™ and it featured a lively discussion of the Nicira deal and its impact on the network vendors. If you missed the feature you can access it at http://youtu.be/BgnnHqtn94k. Feedback is welcome because I want the concept to be as useful as possible. Our conclusion was that while Nicira is no direct threat to networking at all, it does present the risk longer term that IT players will suck up all the features and capabilities that users find useful, particularly for the cloud. This could starve the network for differentiation. The question, of course, is whether network vendors will embrace the SDN goals and make the network the essential place where they’re fulfilled, or watch as the money train departs. That’s a question only time—and the vendors—can answer.