Brilliant skills by the Redmond marketing engine! Yet another set of revenue streams (product, training and support) for M$ doing what you can already do in Linux and Solaris (and various other *nix releases) FOR FREE!

My take is that it's like rolling out a device that can tell you the time on an uninhabited planet. "Congratulations. What else have you got?"

What they should be addressing is the problem the market has in understanding how certain commercial problems can be paralellised by different means to suite this kind of technology.

Although we have seen the emergence of "render walls" and resurgence in ASP servers (no, not active server pages - Application Service Providers) and the like, problem parallellism is a skill set that is incredibly lacking in business and mostly a domain of scientific and mathematic research.

Unless they can make in-roads to helping business to identify opportunities for parallelism in solving their business intelligence questions or otherwise get the most out of the "cluster", (and re-threading their own software maybe this millenium), then it will eventuate to something like Window$ME.

Otherwise, we might be looking at the future of the leased software desktop, served via a cluster that runs your "hired" apps and allows you to see them in a browser on your screen, while they charge for the service by the hour. The market is already there for that service and there are is money it, although heavily offset by massive investment in infrastructure, licensing and expertise.

I guess I am saying that this ain't for the masses.

Thanks for the post Nflight. I know I am coming down hard on their sales hype, but...I AM coming down hard on their sales hype. However, I will definitely be looking into this for my boss and our architect, but I expect that the tie-ins are going to amount to a Corporate Enterprise class Budget :D

An important distinction is that clusters generally require the problem solution to be re-coded to a cluster implementation to utilise the shared memory & disk space features whereas grids lend themselves directly to problems which can be reseolved by the Distributed Computing methodology.

In the mean time, if you are interested in "farming" some of your own machines, try your hand at LTSP for Windows and Linux and you can build your own grid, here now and (in many cases) free.

LTSP was developed mainly for ASP type arrangements, but you can configure it to run the O/S and the Apps locally on the client machines/workstations, with the OS and apps being downloaded from the central server and run locally. Effectively, through some foxy configuration you can craft an environment that amounts to a powerfull grid, yet which is all served from a single machine - which is a DREAM for development and administration.

If clusters still make you drool, Solaris10 is also now free and has a well developed platform that has been doing this for years.

As far as I know there's at least half a dozen or more clustering interfaces to choose from if you look at Linux.

Heck! Even Webmin built in a cluster management interface (although I get the impression this should be called a Grid Management Interface) as a standard feature. If I get a couple more machines I'll look into that some more and give that a spin too.

It's not so much that I don't like M$ (and I don't) it is just that there are so many cheaper working alternatives available already. They want more pie.