Linux: Rebel with a clue

"But you're the InfoWorld Microsoft guy. Why are you at LinuxWorld?"

If I only had a dime for every time someone asked me that last week, I'd be wealthier -- by about 10 cents. But why shouldn't I be at LinuxWorld? After all, I write about the best ways to manage a Windows-centric enterprise network, and we all know that isn't always done by using only Windows-based products.

LinuxWorld was actually an eye-opener for me in a number of ways. For one thing, I couldn't believe the number of Windows XP machines running at this show. And not just in the press room, although here I was mighty surprised to be the only one with Slackware available on his notebook. I was also surprised by the ratio of green hair and lip piercings to khakis and golf shirts; khakis and golf shirts were definitely winning. Except over in the dot.org section where it was still T-shirts and anti-Bill tattoos. Frankly, it was a little disappointing. My first Linux gathering and everyone was corporate. Might as well have been at TechEd .

I saw a couple of upsides to all the Windows machines at the show, however. First, I can stick it to some of those Linux OS zealots I run into from time to time (of which, again, this show had surprisingly few). Second, it shows you really can integrate Linux into a Windows-based installation without compatibility taking a dive and the world coming to an end.

But from a Windows IT management perspective, Linux still isn't a cure-all. The battle for the Linux desktop seemed on many folks' mind at the show, but from my perspective, that was the least interesting aspect. What made my eyes flutter was clustering, blade server management, and storage management.

Microsoft is making a big deal out of its Windows Storage Server product line, along with hardware partners galore. Good stuff, but the combination of IBM's blade server platform running Linux, taking advantage of Linux's embedded server virtualization capability while running a TotalStorage SAN solution on the back end, has so much flexibility that it's mind-boggling. Any server you want, any capacity you want, any number of network interfaces, any amount of fail-over. All in software, all via a sweet little interface with IBM's ultra-quality hardware making you feel warm and fuzzy all over. That is one tough solution to beat.

Scyld Software's Beowulf clusters were similarly impressive, although they were flexible in the opposite way. Whereas server virtualization lets you expand your hardware into as many virtual software nodes as possible, Beowulf clusters let you easily combine your hardware resources into a single, and heavily muscled, logical computing resource. And what a resource. Scyld Beowulf solves many of the problems of the first-generation Beowulf clusters by effectively using a single-system front end. The entire cluster can be completely managed from a single console and looks to that user as a single computer. Even better, groups of nodes can be split off for individual tasks with just a few mouse clicks and can then be added to the grid again to help out with works in progress with the same number of index-finger motions. It's smart, it's seamless, and it really works.