SCO straightens its Linux message

The Santa Cruz Operation says it's likely to announce its plans for a Linux distribution this week. Really, and no kidding this time.

SCO cancelled a press and analyst tour a fortnight ago, after its partners complained that they hadn't seen the script in advance, and could someone please explain what was going on?

News had already seeped out that the company was planning to launch its own SCO-branded Linux distro, leaving SCO flaks in the awkward position of not being able to confirm or deny something they'd just said.

But since this is bound to raise questions about how SCO can juggle four Unixes - OpenServer, UnixWare, its own Linux and the forthcoming Monterey - the fortnight seems to have been well spent on thrashing out a position statement.

And here it is.

"Linux is all about transaction arrival rate," VP of server software John Thomas tells us. "There is a play for Linux in the front-end processing of any business, and that is feeding more machines of UnixWare type clusters that are maintaining state information and running transactions at the back."

That roughly translates as "We can cluster Oracle, and you Linux folks can't", which is fair enough for now.

So how about Monterey, the joint IBM-SCO project to merge AIX, Dynix/ptx and UnixWare on Intel's IA-64?

"Where you need to cache a lot of objects, or need a massive memory footprint to prestate a lot of data for a lot of devices that aren't capable of holding or caching data," says Thomas. Uh-huh... "So servers end up being a memory farm for devices that don't have any."

That's a lot more interesting than it looks for anyone who has been wondering where Itanium chips are going to end up. To date, IA-64 has looked like an answer in search of a problem, with Intel talking up main memory databases of the kind sold by TimesTen and Angora.

But we know that Chipzilla's business model depends on it fabbing and selling an awful lot of IA-64 chips - far more than the number of people who want to run main memory databases.

So the massive server farm idea looks like the latest stop-off on the great Itanic cruise.

The trouble is, there really are only two indisputable advantages of 64-bittyness: a larger number of processes and an increased amount of addressable memory per process, and although we'll buy the first, it's not at all clear to us why you'd need more addressable memory on this hypothetical farm. Suppose a server farm was handling information for a gazillion cellphones. Do you suppose you'd really need to juggle information at any one time in memory than the 2GB+ that a 32-bit OS gives you?

No, the problem for Monterey as well the Project Formerly Known As Trillian (Linux on IA-64) is finding a killer application outside expensive, estoteric niches such as data mining. With Williamette cores set to debut at 1.4GHz, and McKinley not due to ship in
volume until 2002, any OS is that debuts with IA-64 is likely to be met with a collective 'so what?' ®

Related Stories

Itanium prices, specs revealed
Intel's cunning server plans