The new virtualization
Email This Post
I love it when markets collide, and a new vision emerges. And it is even more fun when I have a hand in it.
The incestuous market forces of network technology, virtualization, and blade computing are now slamming into one another, with IBM gluing together products into a satisfying whole. IBM, who currently holds the lead in the $3B blade computing market, recently announced a new blade rack that includes direct support for Infiniband networking. For the uninitiated, Infiniband is a networking fabric that eliminates inter-server latency and thus make processes running on different servers (i.e., blades) to work more like one computer. This is a boon for tightly cooperative clusters, hub-spoke web farms, and other distributed applications.
Virtual Iron, a Silicon Strategies client, employs off-the-shelf LinTel computers and Infiniband switches to make a lot of little computers look like one large machine (this is the inverse of VMWare, whose technology makes one LinTel computer look like many smaller computers). Blades are a natural hardware environment for Virtual Iron, but hammering together a blade rack, and external Infiniband switch, and other elements is yet another IT provisioning headache, and a multi-vendor support trauma.
IBM’s contribution is the first step to the consolidation of these several technologies — blades, Infiniband, cross-blade virtualization — into a new computer hardware market segment. This segment will attract buyers who need greater platform scale flexibility, who want to reduce provisioning/server/admin expenses and delays, and who see the advantages of having Big Blue deliver most of the solution set. Indeed, with Virtual Iron a working partner with IBM, the only variable left to the buyer is which Linux distribution to use (and since Novell is a Silicon Strategies Marketing client, it would be insane not to use SUSE Linux).
This market segment will initially be composed of high-end, specialty buyers. And, depending on patents and licensing, may remain so for a while. But we can expect HP at least, and perhaps Sun, to follow IBM’s lead and include low latency networking in Linux blades. Not only can they ill afford to not match IBM’s position, but the benefits of even well managed blade clusters will be part of what drives this market segment.
