Wednesday, 27 May 2009

Got an itch - the NGS has the thing to scratch it with!

For the more technically minded amongst you, we have an announcement from the team at the NGS RAL site.

STFC/RAL has recently installed a pair of Panasas ActiveStor 6000 shelves, and has made 4TB of its 30TB capacity available as scratch space to the NGS. The Panasas system allows much greater read and write speeds to be achieved than is possible via traditional NFS-based servers –a total of over 1GByte/s has been observed across 24 NGS compute nodes reading and writing simultaneously to the same file system. The system will benefit anyone who needs fast I/O spread across multiple compute nodes.

The file system is available on, and all of its compute nodes, via the /work/scratch directory. First create a directory for yourself (e.g. mkdir /work/scratch/ngsXXXX) and then transfer your working data into it, run your job and finally copy your results out.

Please be aware this area is not backed up, and is liable to period wiping (usually at 1 week’s notice though this can’t be guaranteed), so please be sure to copy your results to safety. It’s also polite to clean-up the file system after you’ve finished with it as this will reduce the number of times we have to wipe it out.

We’d be very interested to hear about any good (or bad) experiences you have with this new system, please feel free to contact us via the NGS helpdesk (

No comments: