System iNetwork

June 19, 2007

Grid Computing: Still Gridlocked for the System i

One of the fun parts about being a products editor is that I get paid to keep my beady little eyes on new technology for the System i without having any responsibility for implementing it at my place of work. It's like pretending to be on a diet, but instead of eating my vegetables, all I really have to do is spoon down the ice cream. But occasionally, despite my relatively devil-may-care position, I still manage to be disappointed with System i technology. And one technology for the System i that definitely disappoints me is grid computing.

IBM launched its grid computing initiative with much fanfare back in 2003. It sounded great then and still does. We don't have to look far to see its promise. For example, there's John Palfreyman's white paper, "Grid Explained," at http://www-1.ibm.com/grid. He separates grid into three levels. Compute Grid, a "starter" level, "offers a way of speeding up a business application by sharing the processing work needed across several computers." Information Grid, an intermediate level, "gives rapid information access to all types and sources of business data from across the enterprise at users' workstations." Enterprise Grid, which Palfreyman describes as the "ultimate form," lets users "execute their most demanding business applications in the minimum time possible and get instant access to all information."

Just taking this at face value, Compute Grid sounds like a great environment for running ERP or manufacturing applications, for instance. Multiple tasks that feed information to each other could run in parallel and come up with answers faster. Manufacturing companies could more quickly figure out complex manufacturing orders that require many materials and actions to assemble the final goods for shipment. And I don't know about you, but Information Grid is just screaming "data warehousing" to me. Remember all those anecdotes about figuring out that diapers sell better when they're stocked next to the beer so busy dads can buy both without looking for either when they're making a run at halftime? What other secrets does your enterprise data have locked away if you could just get the processing power to uncover them? Finally, Enterprise Grid really sounds like having it all, doesn't it? Complex systems of applications running together to propel business activities and keep every user with a need to know informed up to the minute. That could just about be computing paradise from an end user's point of view.

Granted, economies of scale of this magnitude are mainly going to be affordable only for large companies. But those enterprises are out there in our market, some running dozens of System i machines throughout their organizations. Even some medium-sized companies could surely benefit from this technology, too. Harnessing all that capability in more efficient ways is just what grid computing is designed to do. And really, isn't that what the promise of business computing in general is all about: Mobilizing these tools we've invented for ourselves to run our businesses more efficiently? I'd say so, but . . .well . . . grid on the System i . . . it just isn't happening, is it?

OK, back to earth. Sure, when IBM announced this technology, we all expected it would be a little while before it showed up on the System i. And it sort of has shown up, but only sort of. Let's take IBM's own explanation of its "Grid and Grow Express Implementation" program at http://www.-935.ibm.com/services/us/index.wss/offering/its/a1025916, which would seem to be a logical first place to go for information for someone thinking about a grid computing implementation. Except that document states quite plainly that to implement grid, you need to be running Red Hat or Novell SUSE Linux Enterprise, AIX 5L, or Windows. Granted, there are some people running those OSs in LPARs somewhere on their System i machines, but it's not very many yet. So the rest of us are just out of luck?

IBM does offer the IBM Grid Toolbox V3 for Multiplatforms, its "integrated toolkit for creating and hosting grid services," which does support the System i. At http://www-03.ibm.com/grid/solutions/grid_toolbox.shtml, IBM even states that tools for System i support will be available for electronic download as of June 25. That's next Monday.

Or is it? I'd feel a little more confident about that date if it weren't for the facts that IBM refers to this as "support for OS/400 on IBM System i" (shouldn't that be "i5/OS?"), the Grid Toolbox V3 solution brief document to which this page links is copyrighted December 2003, and even the Palfreyman white paper is copyrighted January 2005.

So is this new information or just a forgotten corner of IBM's web site? Whether it's really taken four years for this technology to become available on the System i's native OS, or it's actually been around for a couple of years and nobody's noticed, this certainly seems to be evidence of a lack of emphasis somewhere. And that's unfortunate, because even if there has been a lack of demand in the market for it, grid computing on the System i is a technology that seems to offer some of the best of what computing should be all about.

Posted by at June 19, 2007 10:07 AM