Counting the cost of computer clustering

As some companies’ requirement for faster time to market with engineering developments pushes more work into the virtual and collaborative worlds, IT infrastructures are getting expensive, and management attention needs to turn to ensuring value for money. Large scale computer clustering, with multiple CPUs and network attached storage (NAS), aren’t cheap, and while they’re excellent for topping out IT infrastructures, companies do need to monitor for efficiencies on the one hand, and seek ways to squeeze more throughput on the other. Peter Roberts, alliances director at IT infrastructure analysis tools developer Platform Computing, says IT departments in intensive user sectors, like electronics, high tech, aerospace and automotive, need to be able to provide analysis and management reports. He makes the point that it’s not just a matter of being more accountable, but that analysis shows where improvements to the system can be made, and where additional throughput can be achieved to squeeze more. Hoomi Chadirchi, who heads up engineering infrastructure development at microelectronics giant ARM, says his company has been a big clustering user for around five years. “Management is now focusing on the cost of what is an extensive IT infrastructure,” he says, “so the pressure is on. “We’re like a manufacturing company, but of IP, designing MPUs and selling licenses for others to make. We use clustering to consolidate all the computing resources so we can share data sets – design tools, engineering data, etc. “It’s been very successful: system utilisation is very high; especially if you’ve got lots of jobs, there’s a massive improvement. Engineering testing, for example, can be reduced by more than 100% because you’ve got access to hundreds of virtual computers.” “Two months ago we started to analyse work going through our grid using Platform – looking at it and at the cost of our resources. So for example applications costing £n million to run, who’s using them, when, the utilisation, to help management see where the money is going.” He won’t give hard figures, but says that the engineering and management views the tools provide are helping management decisions and steering a path, not only to accountability, but cost savings, which he reckons are just as achievable for other large scale manufacturing users.