As RISC architecture was to the 1980s, so grid computing is to our time, a fundamental re-arrangement of computing that promises to deliver more speed, and more power, to more people than it would seem Moore's Law might allow.
Open source now has a big dog in the fight. At least that's the the idea behind ActiveGrid, which drew $3 million in capital today, plus big-time VCsMitchell Kertzman of Hummer Winblad and Jean-Louis Gassee of Allegis Capital for its board.>
The idea, according to ActiveGrid's white paper (PDF warning)is that Web Services is creating a "tsunami" of machine-to-machine transactions, and only a new class of computing can hope to keep up with demand.
What kind of transactions? Well, think of links between corporate CRM and ERP systems, giving sales staff immediately visibility of what the calling customer is all about. Or think of RFID, and the instant visibility this gives inventory at every point in the sales cycle. Think of the load identity services are going to place on your servers -- where is Larry now?
Yikes! Yikes, indeed.
The idea is to take the grids that found SETI was not at home, and that now run weather simulations, then put them to use in corporate networks, using open source paradigms and tools. The white paper calls this a "transaction grid," clusters of systems working on-demand.
This is not pie-in-the-sky stuff. Google, Amazon, and Sabre are all based on transaction grids. The idea here is to take these one-off projects and productize them, so more companies and networks can make use of them.
The company's Grid Application Server is all based on LAMP (Linux, Apache, MySQL, PHP, Python & Perl) (I guess it should be LAMPPP), glued together through XML formats like XForms, and scaled to enable what it calls "adaptive transaction processing" -- real-time, context-driven decisions at the point of customer contact.
That's the 30,000 foot view, anyway. Now, who wants to jump up and salute this puppy?