This is the technology that TACC will be deploying this year and putting into production in January 2015:
https://www.tacc.utexas.edu/news/press-releases/2013/wrangler-nsf-grant We'll see how it stacks up for uses in HPC then. On May 16, 2014, at 6:10 AM, John Hearns <[email protected]<mailto:[email protected]>> wrote: http://www.theregister.co.uk/2014/05/16/dis_dssd_disdains_fibre_channel_for_pcie/ Interestign article. Nothing specific that hasn't been seen before - but putting this together at rack scale is very impressive. A petabyte of data available with extremely low latency. I have to say though - the fibrechannel is looking like an increasingly limited technology. The point could be made that fibrechannel has the ability to cope with failure, and multipathing should a controller or HB fai - I wonder how riobust PCIe is to failures. But darn it to heck - we are in high performance computing. My take is that you will tolerate some fragility to get that sort of performance. _______________________________________________ Beowulf mailing list, [email protected]<mailto:[email protected]> sponsored by Penguin Computing To change your subscription (digest mode or unsubscribe) visit http://www.beowulf.org/mailman/listinfo/beowulf
_______________________________________________ Beowulf mailing list, [email protected] sponsored by Penguin Computing To change your subscription (digest mode or unsubscribe) visit http://www.beowulf.org/mailman/listinfo/beowulf
