I have had my knuckles soundly rapped for saying that OpenVMS had heard its deathknell. Since then we have heard that the venerable former-Digital operating system has been successfully ported to Itanium, and so there does seem to be some substance to the roadmap that Hewlett-Packard sketched out for it. And more life to the operating system than I gave it credit for.
I have two things to say in my defence. Firstly, I did say that Compaq sounded the death-knell, and I will accept that since then, HP has given it a reprieve. Compaq's fumbling of the Alpha processor and lack of a migration path for OpenVMS resulted in a rapid decline of the operating system's fortunes, and would certainly have killed it off, if it hadn't been reprieved by Hewlett-Packard which has given OpenVMS users on Alpha a migration path.
Secondly, an operating system that is alive today can still be in terminal decline. The number of OpenVMS servers sold is declining and that means it has a limited lifespan.
Of course, that life could be quite a long one. Operating systems rarely die completely. They carry on for ever, but they shift in vendors' favour. So HP continues to tolerate OpenVMS while it is a money-maker, but prefers Unix and (more particularly) Windows.
This week, I saw a more extreme version of this when I looked at Unisys. The company, which was formed from the merger of Sperry (formerly Univac) and Burroughs, two of the original "Bunch" of mainframe vendors that competed with IBM in the 1970s and early 1980s, still has two mainframe architectures -- but its emphasis has shifted. Skipping Unix, it is pushing hard on Windows.
Unisys' server marketing manager Andy Jordan is a charming fellow who admits to serving time with Burroughs pre-merger, and clearly, in the words of the Unisys anthem, "eats, sleeps and drinks this stuff". He is remarkably upbeat about the company's twin mainframe lines, both of which he says are still selling well. They are at the datacentre level, not the mid-range where OpenVMS exists. The systems have been remodelled so the company can sell them in the same cabinets as its more modern systems. If that sounds familiar, it's what IBM has done with its Z-series mainframes.
And what are Unisys' more modern systems? The ES7000 range, 32-way machines designed to run Windows 2000 Datacenter -- the 'big iron' version of Windows 2000.
Now, anyone with their eyes fixed on the mid-range server market may be vaguely aware of ES7000 as an initiative which apparently failed. Launched in 2000, it was adopted by vendors such as HP, Dell and Compaq, but they never sold any and jumped off, leaving Unisys to soldier on alone.
Our sums led us to believe the company must have sold less than 1,000 of these machines since then, which Jordan cheerfully admits is an overestimate. He is very happy with volumes which would mean disaster in a different market, because this is the datacentre. Not only are the machines very pricey, but those system prices aren't even the main revenue stream. Seventy percent of the income is from services -- turning the product into a tailored datacentre that does everything the customer wants, with workloads managed and responses guaranteed. These are machines that sit in the heart of banks, and each installation is bespoke.
"I'm not surprised partners dropped out," said Jordan. "HP, Compaq and Dell do their business in volume." He expects to see Windows Datacenter flourish, with more emphasis on the product during the Windows 2003 launch. "Now it is 64 bit," he said. "A 64-bit operating system on 64-bit hardware makes a very powerful database server."
According to Jordan, initiatives like UP's Utility Datacenter and IBM's Eliza are efforts to recreate the proprietary environments of old on commodity hardware, and will lock users in just as surely as the old hardware did. He reckons that Unisys is more open because it is using commodity software as well.
And efforts to make Linux suited for the datacentre will inevitably make it into a more proprietary system, he believes: "If I was an IT director, I would be asking what support is in place." Some time down the road, if you put Linux on your big iron, and buy it from IBM, the only support you can get will be from IBM. "Linux may look tactical and departmental, but a vote for IBM's Linux is a vote for IBM, and IBM will use it to lock you in to IBM platforms."
Of course, every vendor will tell you that the others are less open. In his defence, Jordan has a kind of "Goldilocks" argument to make Unisys more palatable. Like the baby bear's porridge, he reckons the company is "just right". Unisys is too small to lock people into its systems, but large enough ($6bn turnover) that you can be reasonably certain that it will still be around once you've signed a stonking big services contract.
Whatever the truth of that one, the message I take away is this. For now, at least, it seems like the datacentre will resist efforts to commoditise it. It's a services sale, and that inevitably means a long-term relationship with your supplier. Whether that is based on trust, or lock-in, at the software or hardware level, it seems unlikely to go away any time soon.