CISS: CISS: The Canadian The Canadian Internetworked Scientific - - PowerPoint PPT Presentation

▶

Jan 15, 2024 28 likes •179 views

CISS: CISS: The Canadian The Canadian Internetworked Scientific Scientific Internetworked Supercomputer Supercomputer Chris Pinchak Pinchak, , Paul Lu Paul Lu , Jonathan Schaeffer, Mark Goldenberg , Jonathan Schaeffer, Mark Goldenberg

SLIDE 1

CISS: CISS: The Canadian The Canadian Internetworked Internetworked Scientific Scientific Supercomputer Supercomputer

Chris Chris Pinchak Pinchak, , Paul Lu Paul Lu, Jonathan Schaeffer, Mark Goldenberg , Jonathan Schaeffer, Mark Goldenberg

Dept. of Computing Science
Dept. of Computing Science

University of Alberta University of Alberta paullu@cs.ualberta.ca paullu@cs.ualberta.ca http://www. http://www.cs cs. .ualberta ualberta.ca/~ .ca/~ciss ciss (for the CISS and Trellis Teams) (for the CISS and Trellis Teams)

SLIDE 2

Thank you to Thank you to… …

n n Wolfgang

Wolfgang J Jä äger ger , , Aiko Huckauf Aiko Huckauf, , Yunjie Xu Yunjie Xu

n n Many, many other people

Many, many other people… …

n n C3.ca, CFI, NSERC,

C3.ca, CFI, NSERC, Netera Netera, CANARIE , CANARIE

n n Dozens of systems administrators and

Dozens of systems administrators and managers from across Canada managers from across Canada

http://www. http://www.cs cs. .ualberta ualberta.ca/~ .ca/~ciss ciss/ /

SLIDE 3

Overview Overview

n n Motivation

Motivation

n n CISS-1, November 4, 2002

CISS-1, November 4, 2002

n n CISS-3, August 2003

CISS-3, August 2003

n n Concluding Remarks

Concluding Remarks

CISS-2 was completed January 2003 CISS-2 was completed January 2003 CISS-3 is scheduled for August 2003 CISS-3 is scheduled for August 2003 Call for Proposals due May 23, 2003 Call for Proposals due May 23, 2003

(Q&A tonight, Monday, 7 PM, Room: Lac Mégantic)

SLIDE 4

Motivation Motivation

Group HPC

Dept. HPC

HPC Centre 1 HPC Centre 2 Server Overlay Metacomputer A Overlay Metacomputer B

SLIDE 5

CISS-1 Participants CISS-1 Participants

SLIDE 6

CISS-1 Resources CISS-1 Resources

8 8 zodiac@UBC zodiac@UBC 20 20 herzberg herzberg@Memorial @Memorial 1,376 1,376 Total Total white@ white@UofM UofM symphony@UNB symphony@UNB stokes@CLUMEQ stokes@CLUMEQ p4-cluster@UofA p4-cluster@UofA myri myri@ @Sherbrooke Sherbrooke monolith@Waterloo monolith@Waterloo minerva minerva@UVIC @UVIC mercury@NRC mercury@NRC maci maci-cluster@

cluster@UofC

UofC Site Site 22 22 32 32 16 16 22 22 96 96 192 192 26 26 236 236 32 32 CPUs CPUs 22 22 hammerhead@ hammerhead@SHARCNET

SHARCNET

2 2 gnome@ gnome@UofS UofS 248 248 driftwood@UBC driftwood@UBC 26 26 deeppurple deeppurple@ @SHARCNET

SHARCNET

8 8 sick kids hospital sick kids hospital 16 16 bugaboo@SFU bugaboo@SFU 128 128 jasper@ jasper@UofA UofA 48 48 aurora@ aurora@UofA UofA 176 176 athlon athlon-cluster@

cluster@UofA

UofA CPUs CPUs Site Site

SLIDE 7

CISS-1 CISS-1

n n November 4, 2002

November 4, 2002

n n At its peak, used 1,376 processors

At its peak, used 1,376 processors … …at 20 facilities at 20 facilities … …in 18 administrative domains in 18 administrative domains … …at 16 universities and institutions at 16 universities and institutions

n n In 24 hours, CISS-1 computed the equivalent of

In 24 hours, CISS-1 computed the equivalent of 3.5 years worth of computational chemistry 3.5 years worth of computational chemistry … …7,593 MOLPRO jobs on November 4 7,593 MOLPRO jobs on November 4 … …over 27,000 jobs in total before and after

ver 27,000 jobs in total before and after

SLIDE 8

CISS-1 Application CISS-1 Application

n n Computational chemistry

Computational chemistry using the MOLPRO using the MOLPRO application application

n n “

“Embarrassingly parallel Embarrassingly parallel” ” (a (a good thing!), capacity good thing!), capacity computing computing

n n Heavy use of temporary

Heavy use of temporary storage storage

n n Developed by

Developed by

Dr. Wolfgang
Dr. Wolfgang J

Jä äger ger and his and his group in the Dept. of group in the Dept. of Chemistry Chemistry

SLIDE 9

“ “Major Major” ” CISS-1 Sites CISS-1 Sites

2:44:34 2:44:34 2919:52:07 2919:52:07 1064.61 1064.61 maci maci-cluster@

cluster@UofC

UofC 2:31:46 2:31:46 3984:29:37 3984:29:37 1575.22 1575.22 bugaboo@SFU bugaboo@SFU 2:36:05 2:36:05 5625:23:44 5625:23:44 2162.47 2162.47 stokes@CLUMEQ stokes@CLUMEQ 10:21:46 10:21:46 4530:30:47 4530:30:47 437.17 437.17 aurora@ aurora@UofA UofA Mean Mean Total Total Number Number Site Site

CISS is primarily about capacity computing. Capability computing is a different (and important) problem. CISS does support MPI and shared-memory threads.

SLIDE 10

Some Lessons from CISS Some Lessons from CISS

n n Social problems are harder

Social problems are harder than technical problems than technical problems

n n Bad technology can make

Bad technology can make social problems worse, but social problems worse, but good technology can rarely good technology can rarely solve them solve them

n n CISS is well-suited for

CISS is well-suited for capacity computing capacity computing

n n Many scientific problems are

Many scientific problems are too resource-intensive for too resource-intensive for SETI@home-style SETI@home-style techniques techniques

n n Social problems

Social problems

n n Exclusive access

Exclusive access

n n Interference with local

Interference with local issues issues

n n Attracting computational

Attracting computational scientists scientists

n n Technical problems

Technical problems

n n Compiler differences

Compiler differences

n n Temporary disk space

Temporary disk space

n n Data movement

Data movement

SLIDE 11

CISS-3, August 2003 CISS-3, August 2003

n n Call for Proposals out now, due May 23

Call for Proposals out now, due May 23

n n During an entire month, we could provide

During an entire month, we could provide 100+ CPU years of total computation 100+ CPU years of total computation

n n Scientific merit is the main selection

Scientific merit is the main selection criteria criteria

We need your help to make CISS-3 a success! We need your help to make CISS-3 a success!

SLIDE 12

Concluding Remarks Concluding Remarks

n n Trellis infrastructure is scalable

Trellis infrastructure is scalable

n n 1,376 placeholders in 18 administrative domains

1,376 placeholders in 18 administrative domains

n n CPUs, memory, storage,

CPUs, memory, storage, software software, , social infrastructure social infrastructure, , and networking are all important and networking are all important

n n The focus is on computational science

The focus is on computational science

n n CISS-1: Computational chemistry

CISS-1: Computational chemistry

n n CISS-2: Molecular dynamics and physics

CISS-2: Molecular dynamics and physics

n n CISS-3 (August 2003): Looking for a large, world-class

CISS-3 (August 2003): Looking for a large, world-class application on the order of 100+ CPU-years worth of application on the order of 100+ CPU-years worth of computation computation

SLIDE 13

Claim to Fame Claim to Fame

n n Jack Van

Jack Van Impe Impe

n n “

“The Bible Prophecy The Bible Prophecy Portal of the Internet Portal of the Internet” ”

n n http://www.jvim.org/

http://www.jvim.org/

n n December 7, 2002 show

December 7, 2002 show (no longer on-line) (no longer on-line)

SLIDE 14

Infamous Infamous

SLIDE 15