WinXP Blade: MS' plan to kill off Linux Web servers
It could work, but they have to think smarter than usual...
There is indeed more than meets the eye to Blade, the WinXP server edition Microsoft introduced as a 'possibility' at WinHEC yesterday. The Register's spies (who seem to be on something of a roll at the moment, and we thank them for that), report that Blade servers are viewed internally as 'cool,' and killer products.
It's what they're designed to kill that's most interesting though - it's Linux. And maybe they're supposed to screw Intel over as well, but we'll get back to that. Blade servers are intended to be cheap, slam-in-and-go boxes that Web hosting operations can just peel off the roll (almost) and shove in whenever they need more server capacity. Now, think about where Linux is really strong, and at the same time think (fairly, mind) about how good Microsoft is at observing, defining and analysing the competition.
Linux is nowhere on the desktop, and not very far at all as a commercial server. But it is pretty compelling as a host for a Web server. It's cheap, robust, flexible, and you can even get nice cheap rackmount units you can just slam in whenever you need one. From the likes of Cobalt, at least until we all got deeply suspicious about them being bought by the other bunch of control freaks at Sun.
Microsoft may frequently be despicable and massively irritating, but it's not dumb. It has obviously (and sometimes publicly) been observing Linux, and has noted its strengths as regards Web servers. Its planners will be keeping an eye on desktop and commercial too, but as we said they're not immediate dangers. The success of Linux on the Web, on the other hand, and Microsoft's failure there, is clearly a big, immediate problem to address.
This is actually where we at The Register live. We run our site on Linux because it's flexible, robust and cheap, and we think we might have to switch to the shifty bunch of control freaks at Sun if we ever grow up. We'd rather stick hot needles in our eyes than go the Microsoft route, and this goes even if Redmond were to give us the code for free, or if we improbably found ourselves in charge of an infinite pile of money. We've gone into this, trust us, this is an entirely commercial decision and has absolutely nothing to do with whether the competing vendors are nice people or not. But they're not. None of them.
Anyway we think our viewpoint goes for a pretty large slice of the market Microsoft will be trying to sell Blade to, in which case it's going to be a tough one to crack, and the level of detail Microsoft's planners have gone into will be important. Cheap is good, but not everything. It's perfectly feasible, given current hardware price levels, for a plug in and go box to be knocked out for not much money at all. But the 'extras' you can't avoid getting into if you go the Microsoft route have to be factored in.
Currently Microsoft products suck you into more Microsoft products, and if you're doing Web stuff and you take your brain out of gear for a couple of seconds you'll likely find yourself running SQL Server, Exchange and goodness knows what else. Blade is simply not going to work if that remains the case - if the Redmond planners are really smart and radical, there'll be decoupling in there too. If buyers can't shove it in without getting lured into a rip and replace of the whole infrastructure, they're not going to be interested. But that is too radical for the Microsoft we know and... well... It's just difficult to conceive of it happening under the current regime, OK?
But it might. And actually, as WinXP is a pretty solid, robust base (there, we said something nice about Microsoft), it could produce a series of highly effective and flexible products if Microsoft could just get over all of the self-imposed, marketing-driven negatives. WinXP really is pretty much plug in and go, provided you're plugging it into the kind of hardware you thought you couldn't possibly need a year ago, but that's dirt cheap now/soon.
So this is a plan that could work, if the planners have thought it right the way through. You could also factor in some of the other hardware-related stuff Microsoft has been doing lately, and come up with possible cunning packaging reasons for Microsoft doing Blade.
For example, there was that Xbox stuff last year when Microsoft improbably claimed that the hardware companies had wanted nothing to do with it, so it had decided to sort out the manufacturing itself. Even though Microsoft doesn't do hardware.* Then just this week we have the Tablet announcement, where the PC companies do get to be involved, but where Microsoft's new friends at Flextronics are listed under "design and manufacturing." So run that past us again - what is it that Compaq's doing?
There's more circumstantial smut. The Tablet release says it'll use "x86-compatible chip architecture," which would make it sound like we're talking about Transmeta here, except that Intel is also listed as a semi technology contributor. You could read that as a clue that Intel might have shouldered its way into the Tablet scheme late in the day, and they didn't update all of the details. Just like they didn't when Intel blew AMD out of the Xbox deal.
But we digress. There is, in WinXP, support for Transmeta Crusoe and for AMD 64 (a lengthy portion in the DDK, in the latter case). Microsoft is clearly playing footsie with Intel's mortal foes, and we wouldn't be in the slightest bit surprised if Transmeta's server offering and/or AMD's were adopted for a Blade reference design. Nor, on the other hand, would be surprised if Intel managed to gazump both of these at the last minute. That's obviously one of the points of footsie, and if it worked it will obviously have cost Intel.
But we would be surprised if Blade didn't turn out to be another instance of the company that doesn't build hardware defining the hardware and software platform, serving it up to the OEMs and kicking its friends at Flextronics loose on it as well. Go look at the message accompanying the joint Microsoft-Intel PC2001 spec, where it says this is the end of the series, and we'll be doing our own things from now on. This, too, is significant. ®
* Microsoft keeps saying these days that it doesn't do hardware. When the PR people insert such things they're unconsciously telling you they are worried about you worrying about the things they say they're not doing. In 1984 (no significance in date) The Register owned a piece of hardware you put in an IBM PC. It was a memory expansion board. A Microsoft memory expansion board.
Sponsored: Hyper-scale data management