NCSA RP Update John Towns. NCSA Resource updates Cobalt –CXFS update Lincoln –production since...

8
NCSA RP Update John Towns

description

Storage Updates Lustre-WAN from Indiana mounted –mounted in 2 of 4 login nodes on Abe for testing –after another 2weeks of good testing, will mount on all 4 login nodes Uberftp –released bug fixed and large feature release recursive directory support for all commands Uberfs released –ssh and all other off the shelf data transfer supported filezilla, rcp for archive system

Transcript of NCSA RP Update John Towns. NCSA Resource updates Cobalt –CXFS update Lincoln –production since...

Page 1: NCSA RP Update John Towns. NCSA Resource updates Cobalt –CXFS update Lincoln –production since mid-March –final configuration 192 compute nodes – Dell.

NCSA RP Update

John Towns

Page 2: NCSA RP Update John Towns. NCSA Resource updates Cobalt –CXFS update Lincoln –production since mid-March –final configuration 192 compute nodes – Dell.

NCSA Resource updates• Cobalt

– CXFS update• Lincoln

– production since mid-March– final configuration

• 192 compute nodes – Dell PE 1950iii dual, quad-core Harpertown

–16GB memory • 96 S1070 Tesla units from NVIDIA

–double precision support –~345 GF peak double precision per S1070

–~2 TF peak single precision per S1070

“Using Replica Exchange MD effectively allows use of the entire Lincoln cluster - and at this point, 96 nodes of Lincoln (48 Tesla units) is equivalent to 50 frames of BG/L… the machine not only is a useful resource, but it also provides an inspiration to think about large scale MD simulations and how to exploit parallelism very differently. We now have several ideas on the table that may help to attack "impossible" problems in the future.” -- Axel Kohlmeyer, U of Pennsylvania

Page 3: NCSA RP Update John Towns. NCSA Resource updates Cobalt –CXFS update Lincoln –production since mid-March –final configuration 192 compute nodes – Dell.

Storage Updates• Lustre-WAN from Indiana mounted

– mounted in 2 of 4 login nodes on Abe for testing– after another 2weeks of good testing, will mount

on all 4 login nodes• Uberftp

– released bug fixed and large feature release • recursive directory support for all commands

• Uberfs released– ssh and all other off the shelf data transfer

supported• filezilla, rcp for archive system

Page 4: NCSA RP Update John Towns. NCSA Resource updates Cobalt –CXFS update Lincoln –production since mid-March –final configuration 192 compute nodes – Dell.

Training• CI Tutor very active:

– 1250 enrollments in the 24 courses offered during the first quarter of 2009

• Several CI-Tutor tutorials in progress– “Getting Started on the TeraGrid”

• finishing up– “Introduction to accelerator technologies”

• started developing short tutorial– “Using BigSim to Simulate Petaflops Computers”

• almost completed• Blue Waters item which could also have some usefulness for the TeraGrid

• TG’09 involvement– Sandie Kappes giving presentation on CI-Tutor– Masoud Sadjadi (FIU) giving on talk on his pathways fellowship that uses

CI-Tutor • finding it very useful for his course in conjunction with providing access to the UI's Elluminate web-conferencing software to conduct his class

– Galen Arnold will give short presentation on student day on how to setup a software environment with MPI, OpenMP, etc on a desktop/laptop that can be used to run the code given in the exercises within CI-Tutor• will also add an area within CI-Tutor with the information

Page 5: NCSA RP Update John Towns. NCSA Resource updates Cobalt –CXFS update Lincoln –production since mid-March –final configuration 192 compute nodes – Dell.

Successful Deployment and Test for Gateway User Count at NCSA

• Science Gateways Capability Kit deployed on Abe and Mercury

• GISolve Gateway submitted test jobs with attributes to Abe

• Integration with AMIE accounting process underway

GridShibfor GT

WS GRAM Service

Logs

Java WS Container(with GridShib for GT)

Abe

GRAMAudit Table

AMIEupload

TGCDB

GISolve

Page 6: NCSA RP Update John Towns. NCSA Resource updates Cobalt –CXFS update Lincoln –production since mid-March –final configuration 192 compute nodes – Dell.

POINT (NSF SDCI) Highlights• PAPI, TAU, PerfSuite, Scalasca available on

TeraGrid platforms, support/maintenance ongoing• TAU integration with Charm++ “Projections”

performance framework• Tutorial sessions on performance engineering

and tools at HPC conferences:– SC ‘08– LCI ‘09– ICCS ’09 (with European VI-HPS project)– TG ’09 (with NSF IPM SDCI project)

Page 7: NCSA RP Update John Towns. NCSA Resource updates Cobalt –CXFS update Lincoln –production since mid-March –final configuration 192 compute nodes – Dell.

TeraGrid 09• June 22-25, Crystal City Hyatt, Arlington Va

– http://www.teragrid.org/tg09– hotel block is now closed, but you can still get the rate on

available rooms!– registration at ~350 and still growing

• Complete agenda is posted to the web site– adjustments/tweaks seem to have died down

• Proposal funded for student participation– ~$120,000 in student participation support– ~120 students participating!!

• Economy has impacted vendor sponsorship– still trying to garner a couple more sponsors

Page 8: NCSA RP Update John Towns. NCSA Resource updates Cobalt –CXFS update Lincoln –production since mid-March –final configuration 192 compute nodes – Dell.

TeraGrid 09• Program highlights

– Ed Seidel, Paul Avery and Tom Cheatham keynotes– Daily user-focused sessions

• Tutorials all day on Monday• SAB Roundtable• Agency Roundtable• XD Transition discussion• joint XD Requirements BoF

– Science, technology and education tracks• ~24 science, tech and EOT slots

– Working groups and BOFs through the week– Significant student program

• programming contest• paper and poster competitions