Research Computing Panel CSG Seattle '02

1
Research Computing PanelCSG Seattle '02

• Research Computing Support
• at
• Penn State

2
Penn State

• 82K students
• 4K faculty, 10K staff
• 24 locations
• PA's land grant

3
Research at Penn State

• 20 Colleges, Consortia, etc.
• 500M research annually (out of 2.4B total)
• Navy, HHS, NSF, NASA etc. make up about half
• Industry, state, etc. make up the rest
• Near term areas of emphasis/strength
• Life sciences
• Children, Youth and Families Consortium
• Materials

4
Goal

• Fill in gaps in research enterprise to
• Make research easier to do
• Make research less expensive to do
• Reduce time to discovery
• Take risks where faculty can't or won't
• Lead by example quickly or get out of the way
  even more quickly

5
Background

• As with everything else on campus - research,
  research computing, and research computing
  support is a highly distributed venture with an
  amazingly mobile and changing set of spheres of
  influence
• Fine-grained (PI level) and profound independence

6
Research Computing Support at PSU

• Software
• Hardware
• Consulting

7
Software

• Site license mgt for appropriate software
• Roots in site licensing for various unix Oses
• Includes titles such as SAS, SPSS, Labview,
  NASTRAN, FIDAP, IDEAS, GAUSSIAN, Mathematica,
  Matlab, NAG, IMSL, etc.
• We'll seed new acquisitions and help market their
  usefulness
• Software on our on resources that might fall out
  of this

8
Software

• Bioinformatics portal - collection of genomics
  tools in one place
• ETD support

9
Hardware

• IBM SP - seeing limited use at this point
• Large memory cluster- AIX based
• Intel clusters
• Lion-X (32x2 500MHz PIII)
• Myrinet, GigE, 100Mb
• Lion-XE (112x2 1GHz PIII)
• Dolphin, 100Mb
• Lion-XL (80x2 2.4 Ghz PIII)
• Quadrics,100Mb Bluearc

10
Hardware (?)

• OpenPBS
• Maui scheduler consortium member
• Compilers du jour, numerical libraries
• All hardware acquired via consortium building
• We make a technology decision considering many
  factors, and get others to buy in. In return,
  they share unused cycles. come from lots
  places NSF MRI, SCREMS, IGERT, etc.
• Also run a co-location center

11
Hardware (?)

• Visualization
• Immersadesk
• Immersive Environments Lab
• Plotting
• Outside of central IT
• Clusters
• 10-15 small to large clusters out there (16 to
  100 nodes)
• We'll help where we can and share all we know
• Many instances of less than efficient operation
  and performance

12
Hardware (?)

• Outside of central IT
• Clusters
• 10-20 small to large clusters out there (16 to
  100 nodes)
• We'll help where we can and share all we know
• Many instances of less than efficient operation
  and performance
• Visualization
• 3 RAVEs
• C3 (CAVE) in Applied Research Lab

13
Hardware

• Partnerships
• Level 2 iVDGL site
• Tobacco
• Meteorology, Math, Stat, Astrophysics, etc.

14
Consulting

• Distributed workstation sys admin - cost
  recovered
• We'll help you build a cluster but will encourage
  you to see the benefits in partnering with us
• We'll help you port codes to a dist. Mem. Par.
  computing model
• Statistical consulting done elsewhere - Stat,
  Social Science Research Institute, Meth. Ctr.

15
Grid

• Hypereality ratio is heading back towards one
  instead of infinity which is a good thing
• We'll try to put a man on on Mars too but if we
  wind up only getting to the Moon, that'll be OK
  and still very cool and useful.
• As already mentioned, iVDGL