HPCclustersinResearchComputing.ppt

HPC clusters in Research Computing HPC (cluster) Services offered by Research Computing to the Partners Research Community Dennis Gurgul Jan 25th 2007 The Services Linux based computing clusters for large-scale scientific computation or data analysis Offer computing power to Partners research community and collaborators About 150 accounts, constantly active accounts 20-30 Average 300-500 jobs submissions per day  User support: Non-standard applications support : Open source software , in-house developed applications …etc User guidance for system access and job submission, trouble shooting, performance analyzing. Application/system integration with HPC Beyond our cluster, general HPC support. E.g. Proteomics, DNA sequencing data processing.. Brief History -- Intro-how did we get started (GCG) email P “Lets build a cluster see what happens” —Martinson, Jon 2002 Brief History -- Intro Research Computing Infrastructure Brent Richter Dennis Gurgul Steve Roylance Jerry Xu Aaron Zschau HPCGG, RPDR, I2B2, *NIX, CLUSTERS Brief History -- Intro Agenda: Intro Resources Detailed Technical Description (Hardware/OS/etc) Apps / Users Success Stories On-going Changes Future Plans How to Get an Account Computational Resources 61 nodes, LSF, 8 TB temp storage, AMD/Opteron 25 nodes, SGE, 0.5 TB temp storage, Intel 32 bit (oldest – Matlab) 15 nodes, LSF, 0.8 TB temp storage, AMD/Opteron System Descriptions: hpres Deployed 4th Q 04 HP XC4000 cluster mgt. package (de-branded RHEL), which includes LSF, Two HP Proliant DL585 login nodes w 2 AMD/Opteron, 4 GB RAM, two mirrored 72 G SCSI One HP ProLiant DL585 head node 61 HP ProLiant DL145 compute nodes w 2 AMD/Opteron, 4 GB RAM, 160 GB ATA HP MSA1500 SAN w 42 300 GB SCSI as 8 .8 TB LUNS, ADG, 3 hot spares over fibre HBA to login nodes NFS over GB Enet via HP ProCurve switches. HPRES Architecture System Descriptions: rescluster2 Deployed in 4th Q 2003 Rocks with included SGE on CentOS Dell 2650 PowerE