"Fabric Based Computing" the future of Fabric Apps?

If you attended the Gartner Data Center conference a week ago, you may have already heard the buzz on “Fabric-based computing”. It’s a fancy new label for Cisco’s entry into the market with the UCS platform. It’s almost a marketspace of one, but with the addition of the HP BladeCenter matrix and a couple other smaller players, 3 leaf networks and Liquid Computing. And that surprisingly, 30% of Global 2000 companies will run some form of it by the end of 2012.

Is this really a market? Or just a server bundling exercise being wrapped in revisionist PR. When Gartner first published about Computing Fabrics they were speaking to the emerging enterprise compute-clustering space, now populated with just a few remaining players such as Egenera, Scalent and Surgient. The others have died or been bought – some just for their IP – such as Virtual Iron/Oracle, Cassatt/CA, Platespin/Novell, Qlusters. The problem is that the Cisco modular blade server / network switch chassis is a world away from an offering that does virtualized server compute, shared system memory, automated/dynamic resource management, etc.

Also, I’d love to think that “Fabric-based computing” will enable a rebirth of the network-based application space. Again, products like UCS are just a server chassis (Cisco now offers both blades and rack-mounted servers) that run Intel processors and standard operating systems. This is decidedly different from how Invista or RecoverPoint runs actually within the network fabric, able to work with heterogeneous host operating systems. Not saying it’s better or worse, but just that it’s different.

Anyway, in Gartner’s study on the topic (just published, not yet available online), they point out that “Adoption will likely follow a similar pattern to that of the early years of the blade server market.” This is because UCS is primarily a blade server. No need for a fancy new name. And no need to invoke the spirits of Fabric Application or Compute clustering.