I thought the same thing about this article. Cassatt was started to produce an "N1" type of software, creating a utility grid of horizontally scaled Linux boxes. Now Cassatt has reinvented itself as a provider of tools to manage consolidated applications running on VMware on x86 servers.

Now think about that. Cassatt started as a "macrovirtualization" play, that is making many computers work as if they are a single computer. But now, it simply adds a layered management function to a "microvirtualization", that is a single computer working as if it is many computers.

The common aspect of both Cassatt visions is provisioning and reprovisioning servers. That is really all Cassatt does now. But that is a market with many players: Sun N1, HP Insight Manager, IBM Director, Opsware, Cisco VFrame, etc.

Of course, the real reinvention was not Cassatt, it was Virtual Iron, which proposed to create a "virtual SMP", not a grid, from cheap x86 Linux boxes. What I could not figure out, was Virtual Iron going to create a Sequent Dynix like NUMA machine, or a VAX like shared system image cluster?

Who knows. We do know Virtual Iron now simply provides a shrinkwrapped version of Xen and some layered management tools.

Now here is what is really interesting in all of this. Sun tried to do a VAX like SSI clustering OS years ago (http://research.sun.com/techrep/1995/abstract-48.html). And the motivation to do this was the predicted emergence of Java based distributed computing, the same Java grid Cassatt started chasing five or six years later. But Solaris MC, in a lesser, non SSI form became Sun Cluster 3, and the reason for Solaris MC (the Java grid, what Yousef Khalidi called "disaggregated computing") became irrelevant as the key J2EE appservers provided their own clustering mechanisms, web servers didn't need clustering mechanisms (just a network load balancer in the front with NFS at the back), and Oracle started making OPS actually work as advertised.

Despite this realization at Sun in the late 1990s that a multicomputer OS was not needed, some continued to pursue the windmill of the Java grid as if it were a dragon to slay, Rob Gingell among them.

In his defense, perhaps Gingell thought Linux was the key difference, or as open source infiltrated middleware a vendor of a Java grid layer would become the key value adder.

Pragmatism has won, and management, albeit boring, is the real problem which needed to be solved all along.

But the lesson is beware the latest trendy thing. Inertia is the fat son of the athletic father called momentum. Which is why pragmatism usually wins.

Posted by Mark on September 29, 2006 at 07:20 PM PDT #