After spending the past four months in close proximity and constant communication with the tech journalists and analysts covering grid computing, the next generation data center and utility computing, I still wonder at the major misconceptions that are purpetuated (perpetrated?). Here's one of them.
Most of the article sets out the list of failed business deals that IBM's engaged in regarding various forms of outsourcing and "on demand" services. The author (Dan Lyons) then makes a statement like the one below. It bespeaks two things: the superficial nature of tech journalism, and the inability of the technology providers to make simple the distinctions among approaches being lumped together as "grid" or "on demand."
On the first issue -- tech journalism superficiality. Tech journalists are called upon to learn an awful lot, break it down to its essence, package it for a wide variety of readers, hit the deadlines, and ... oh, by the way ... manage to do this simultaneously for six or eight different topics which change weekly. It's no wonder that when I read a journalist's piece about something with which I'm involved or something I've witnessed, I marvel at either the inaccuracy or the lack of depth.
On the second issue, "on demand" and the complete delegation of resource manangement to an outside agency... it ain't so. That model MIGHT well describe the days of the timesharing bureaus and centralized glasshouse models of the 70's and early 80's, but not necessarily today's On Demand models. The point of Grid computing (... OK, ONE of the points of Grid computing ...) done using service oriented infrastructure and late binding is that compute, storage and communication resources can be managed in a federated manner. The "customer" in Dan Lyon's scenario does not have to turn over complete management responsibility to the "service provider". Sets of resources can be offered up to the customer on an as-needed basis, federated in such a way that the policy defining the shared use of the resource defines how a particular computer, storage unit, or data set is utilized, under what conditions management control is transferred (and for how long), with the assurances that the customer's requirements can be well-defined in advance and the performance of the service is auditable.
Why is IBM's On Demand computing efforts missing the mark? Because they're not yet based on shared infrastructure that supports this kind of federation of resources. This stuff takes time. One can possibly fault Palmisano and crew for mistaking a clear vision for proximity. On Demand IS doable, but there's some heavy lifting yet to be done to build out the basis on which it can be done successfully.
Blanket statements like Dan Lyon's serve to just scare the customer and the industry, and muddy the issues. It does us all a disservice. The On Demand vision demands more studiousness by the reader and the tech journalist, and better explanation and clarification by the vendor and supplier.
Link: Forbes.com: IBM's Growth Engine Sputters.
The On Demand approach relies on customers surrendering all control over tech, but some clients are deciding to take it back.
