IBM's goes back to the mainframe

One of the big risks when announcing a new green business initiative is the fact that it can simply serve to highlight the things you were doing wrong in the first place.

Back when the supermarkets first started stocking fair trade products I remembered thinking that, by definition, everything else they stocked should now be labelled unfair trade. Now, every time a company releases a press release claiming they are now giving out 30 percent fewer plastic bags or have cut their lighting bills or reduced their product packaging the cynical part of me wants to say "why the hell were you handing out so many bags/leaving so many lights on/packaging everything in 30 sheets of cling film in the first place?"

I was reminded of this last week when talking to Steve Bawden, CTO at IBM's server and storage division about the company's hugely admirable project to slash its datacentre energy consumption by 80 percent by consolidating around 3,900 of its servers onto 30 mainframes.

The question that sprung to mind was if these System z mainframes were so great - and IBM has sounded as enthusiastic as a South American football commentator when talking them up – then how did the company get into a situation where it was running the best part of 4,000 servers in the first place? Why has it taken until now for IBM to realise the most efficient way of setting up its infrastructure is based around a technology that it has had since before I was born?

Bawden's response was that IBM followed the same model of many other firms whereby individual departments had their own budgets and were left to invest in their own systems. Over time this led to a highly distributed infrastructure, which may have meant each department faced lower upfront costs for their IT but also proved extremely inefficient in terms of energy use and management costs.

This inefficiency was tolerated because as long as energy costs remained low the inherent profligacy of running thousands of servers with low rates of utilisation did minimal damage to the bottom line even while it did significant damage to the environment.

Even when energy and management costs did begin to rise it was less than easy to transfer across - or should that be back? - onto mainframes because major projects had to be undertaken to make the various applications portable.

IBM maintains that much of this work has now been done and as a result the consolidation of its 3,900 servers onto 30 System z mainframes should, in the words of Bawden, prove "relatively simple".

Of course it is slightly unfair to dwell on how IBM got itself into such an inefficient scenario when it is now committed to resolving the problem. But it does serve to illustrate quite how far down the list of priorities energy conservation has been over the past few decades and how horrendously inefficient the IT infrastructure of even the most tech savvy companies has become as a result.

Where IBM is to be praised is in its recognition of these facts and its willingness to try and solve the problem by returning to a technology that many regard as something of a dinosaur. Should the company prove successful in its consolidation project, and there is no reason to suspect otherwise, then it might just pave the way for an industry wide swing away from sprawling datacentres and back towards the mainframe.

After all it is a technology that, as Bawden observes, is significantly more energy efficient and easier to manage than a never-ending sea of servers. "It's like with transport," he muses. "It's always going to be greener for people to get on the train rather than drive into work on hundreds of mopeds."