Fujitsu wins £15m Welsh supercomputer bid

Probably not a single supercomputer.

That seems like an awful lot of hardware for the budget, and the dispersed nature means that it is far more likely to act as a group of smaller clusters that talk together than a single super-computer.

You will never be able to drive the WAN links at sufficient speed to spread anything other than encapsulated data type problems (like SETI@Home but larger) to the remote sites.

I would be interested in seeing how the power spreads around the six sites, because although the total amount of compute power may seem high, chances are that the power in any single part of the environment will be a fraction significantly under half of the total. That should put it much further down the top 500 that the 'low 30's' quoted in the article.

Also, by the time it is delivered, there will be new systems springing up in China, USA, Germany and even the UK!

Oh well. I have contracted for Fujitsu in South Wales before. Maybe I ought to dust off the old CV. Might be interesting to do some non-AIX work for a while, and I now have Infiniband experience.