Dr Arthur Trew EPCC Director A Research Computing Infrastructure for Edinburgh
who are we building this for? local researchers, eg. –Computational: UoE HPC service, BlueGene/L, BlueDwarf … –Data: ScotGrid, QCDGrid, eDIKT, … national research consortia –Computational: QCDOC, HPCx, DEISA –Data: NDCC visitors –Computational: HPC-Europa visitor programme –Data: eSI … new projects/bids –HECToR, Brain Imaging, systems biology … –next generation machine design …
what’s the problem? we thought it was networking and CPU … but, a survey of users added data … BIG TIME!... and this projection only included projects with secured funding TB
in summary while individual needs varied all required some mixture of –more CPU/memory –more data storage –faster access researchers were used to buying their own CPU servers so, our strategy was to complement this by providing a support infrastructure Stage 1: we used £1.68 M of SRIF1 to: –create a fast, research network in parallel with edlan –install a research data storage facility
SRIF KB SRIF AT The last mile: SRIF1 JANET BAR Royal Edinburgh Hospital Library Appleton Tower Kings Buildings Holyrood Robson Building EaStMan Router Old College Western General Pollock Halls New College RESNET Sick Children’s Hospital Medical School Little France BUSH 2Mbit/s 100Mbit/s 1000Mbit/s
… and data too 155 TB SAN + 36 TB tape backup –available to all e-science researchers total investment SRIF1 + project funds: £2.2M
short of space … but all the new facilities could not be fitted in the JCMB machine room £4.2M from SRIF2 to refurbish a new research computing facility
SRIF KB SRIF AT SRIF KB SRIF AT ACF 10 Gb/s The last mile … Phase 2 JANET BAR Royal Edinburgh Hospital Library Appleton Tower Kings Buildings Holyrood Robson Building EaStMan Router Old College Western General Pollock Halls New College RESNET Sick Children’s Hospital Medical School Little France BUSH 2Mbit/s 100Mbit/s 1000Mbit/s
the research SAN KB AT SRIF/ SJ4 Fibrechannel 10 Gb/s HPCx iSCSI NAS QCDOCBG/L SAN
what facilities are available to me? CPU –UoE HPC service (lomond) – 52-pe Sun E15000 –BlueGene/L (bluesky) – 2,000-pe IBM R&D machine –capability computing for the initiated data –research SAN – 155TB (disk), 36TB (tape) Sun 6290 networking –SRIF network secure machine room space … me with your needs and we’ll see what we can do other facilities by arrangement with project owners –BlueDwarf, ScotGrid,
a free lunch? the ACF is a strategic University facility –and hence open to all researchers although a real capital asset, the recurrent support for –facilities management –maintenance, power, space charges … –and, perhaps, user support and applications porting/tuning … has to come from project funds what is it you want/need?