Utility data centers--IT utopia?

The idea of the utility data center is to make IT a resource that can be turned on at will and used by many applications. The chief information officer (CIO, or IT manager) can adjust the resources for each one up and down at will.

HP uses the utility data center name, and IBM offers "autonomic computing." Sun Microsystems has started talking about its own proposal, N1.

Recently in London, Sun showed me a mock-up of what its N1 data center could ultimately look like. An IT manager gets a request to double the capacity of an online ordering system so it can handle 300,000 customers during the Christmas rush (Christmas rush? That alone is enough to assure you this is a hypothetical scenario!), and manages the whole thing from a graphical console.

First, he calls the application up, and is shown a screen that tells him the current performance guarantees, and the resources required. He types the new number of users into a box. The system thinks a while, then tells him there are enough free resources to support the new service level.

The system then works out what is required, and quotes an extra $10,000 per year as the charge-back cost for the IT manager to pass on to the user department whose application is being boosted.

A final click accepts the changes, and the N1 data center applies them. It "soft cables" a few extra blade servers and an extra Web server together, and puts the application across them.

All this, we are told, will happen without the IT manager having to rewire or reprogram anything. This is, of course, a massive saving in IT staff. Normally, one admin can handle about 15 servers, and by 2005, N1 will let him or her handle more than 500. Currently, the admin can handle a terabyte of storage, and N1 will boost that to 100. This represents a lot of staff savings--so utility data centers are bad news for system administrators.

For the IT manager, this is a utopian ideal. Of course, even his or her job looks like it's being automated into a shadow of its former self. If N1 can work out chargebacks, it must have a sophisticated understanding of company politics, and what user departments are prepared to weather.

Even Sun admits that it won't happen till at least 2005. Perhaps that is when Sun expects us to have another Christmas consumer boom. But there is a drawback. Even in 2005, this Sun proposal is a Solaris-only environment. Which seems very, very strange, given the fact that for the last umpteen years, most vendors--and Sun in particular--have been promising "open systems," which link multiple vendors' kits together.

We all know that Sun Microsystems (like all of the others), for all its talk about open systems, would grab a monopoly with both hands given a chance. Those who remember the start of the Unix wars, when Sun attempted to mandate its SPARC processor for System 5.4, will know that it is only the lack of opportunity that has prevented Sun being a monopolist on the scale of Microsoft.

But why be so upfront about it? There are two reasons. Firstly, at the moment, multi-vendor data centers do exist, and many of them include Sun's kit. However, the big costs in those multi-vendor shops, even after years of open systems development, are in the integration--whether that is done by in-house admins, or by service providers and services agencies. Those costs are both a big part of IT expenditure and a big revenue opportunity for a vendor with a services arm.

And Sun, it is often observed, has no services arm to speak of. HP and IBM have. So multi-vendor shops represent a great revenue source for HP and IBM--especially if the utility data center concept is not yet a reality. HP and IBM can promise a utility data center now and, if it isn't easy to put together, they can still make money by putting a lot of staff onto the case in exchange for a large check.

Sun has to sweat it until the thing actually works. And even then, it can only make a go of the utility concept by selling it to users as a single-vendor solution. "Take our data center," they will have to say. "It is technically an open system, but you have to buy it all from a single vendor." And the important bit: "You will get such big savings from it that you won't mind this fact."

The original open systems movement arose from dissatisfaction with IBM's ability to shut everyone else out of data centers by owning all the interfaces. During the 1990s, its focus has shifted to Microsoft's efforts to gain market share with monolithic solutions.

Now in the 21st century, one of the pioneers of the open systems movement is trying to lock users in, in exchange for cheaper cost of ownership. Cheaper, more flexible IT might justify yoking yourself to a single supplier, but we will have to have the pitch explained very clearly and honestly.

And for data center sales, a clear, honest sales pitch could well be a first.

Would you consider a utility-based data center to help your company reduce costs? TalkBack below or e-mail us.