[Bush2000]

IBM's mainframe business is making money. And no sign of it stopping.

So is Dell. What's your point? I didn't say its business was dead. I said that it is a mature business that isn't growing.

And some of the new mainframe deployments (new deployments? Yes, sorry to wreck your dreams) are using Linux. And replacing their existing NT with it.

Not many, dude. Moving to untested Linux from battle-hardened VMS/CICS/MVS is a fool's errand. I don't know of any company that's doing it seriously.

But no, the military keeps buying bigger and bigger mainframes and writing new software for them.

First, they're deploying old software to those mainframes. And secondly, since you're using that hardware, it doesn't make sense to use clustered PCs because you've already made the investment in mainframes. So you write new software targeted at those new mainframes. No mystery there. And it doesn't make the mainframe approach any better. Just easier to manage and deploy based on past needs.

Yeah, there isn't any new technology coming out of universities. Let me see, AI, genetics, astronomy, and physics. Nope, no advancements there.

I hate to be the one to break this to you but ... the reason they're using that hardware is that they don't have the budget to invest in new hardware, dude.

Seems like the Windows crowd wants it both ways. Clustering hasn't had the chance to displace mainframes because it's so new. (Never mind that Cray was using clustering in the 70's) But Windows is established in the desktop market and Linux just it too new to displace it.

Windows isn't the only way to do clustering and I've never argued that. You can cluster mainframes, for chrissake. But clearly, this is the first time that that kind of power has been available at such a reasonable price level. Whether Linux or Windows, there's no arguing that clustering is going to give mainframe vendors a serious run for their money in the future. IBM wins because they play both sides of the fence. If you want to bet on a winner, bet on IBM.

[Knitebane]

Not many, dude. Moving to untested Linux from battle-hardened VMS/CICS/MVS is a fool's errand. I don't know of any company that's doing it seriously.

Then you didn't even look at the link. That bank is doing it, and seriously. And it's a new mainframe installation. When I did a Google search (I know, evil Google, powered by Linux) I came up with at least a dozen new customers using mainframes. IBM is dumping money into R&D for it's mainframe business. As IBM generally cuts the budgets of divisions that aren't growing, seems like mainframes are a growing market for them. Guess they know something that you don't.

And to talk to the other side of your schizophrenic brain, moving from untested Windows from battle-hardened Unix/MVS was a fool's errand. But a lot of companies did it and a lot of Windows weenies cheered them on. Poor fools. A lot of them are switching back, poorer in the pocket but richer in wisdom.

First, they're deploying old software to those mainframes.

How, exactly, does one use old software for new projects?

So you write new software targeted at those new mainframes.

Uh....yeah, that's what I said. So which was it, are they deploying old software or writing new software for the new mainframes they are buying? And they are buying them. DARPA just bought a huge new system. A Cray, I think. The Air Force is famous for buying new mainframes on a regular basis.

I hate to be the one to break this to you but ... the reason they're using that hardware is that they don't have the budget to invest in new hardware, dude.

Not quite, dude. They are writing new software for some of their old mainframes. And they are buying new mainframes, too. The investment is in experience and they know better than to try to switch to Windows machines after a few well publicized failures. The software they wrote three years ago will run on the machines they bought last week, and the software that they wrote this week will run on the the machines that they buy three years from now. That's not true of Windows machines, no matter how clustered, and any Windows sales rep that tells you that it will be true if you buy today is a damned liar.

But clearly, this is the first time that that kind of power has been available at such a reasonable price level.

That remains to be seen. Purchase price plus TCO for a cluster may or may not be better than purchase price plus TCO for a small mainframe or mini. Intel clusters are built for the same reason that RAID systems are built. You are building expecting failure. Building a replacement machine and reinstalling software (or paying for a service contract so that someone else can do it) will get expensive. PC clusters are certainly cheaper to buy than a mainframe. Whether you acutally get more bang for your buck once you factor in maintenance is not yet known. Replacing "Old Reliable" with a PC cluster may well be a fool's errand, and some of those new mainframe purchasers have been there with Novell to Windows "upgrades."

Mainframe clusters are built strictly for performance reasons and have the pricetag to prove it. PC clusters are built for performance on the cheap. There are places where that makes sense. Replacing your existing mainframe with one aint it.

Remember: Cheap, fast, reliable. Pick two.