Sunday, 22 March 2009
Our Research Computing group spent some time last week discussing how we finance our research computing - currently we fund a core service centrally. Departments can then add to this or buy dedicated time from specific research grants. What we don't do is charge departments for all of their use of our High Performance Computing - my view is that it would be too much of an administrative overhead, and we should be providing some sort of pump priming service, as well the opportunity for undergraduates to use it. However, we do need a simple model for calculating costs both for fEC and so that we can recover the true costs from research grants. One of the biggest costs associated with these machines is power and cooling. They need a lot of power - a lot of which they seem to turn into heat - which then needs removing. It's causing problems across the sector, and has led to a number of institutions looking to see if we can share data centres for this sort of specialised computing, with properly designed power and cooling facilities. Peter Tinson of UCISA blogged about a number of projects recently, and summarised the situation better than I could, so I won't try and repeat it here. We're involved in two of them - YHMAN and SHED - and are waiting with interest to see how they progress.