Big-FE-teragrid-abstract: Difference between revisions

From Earlham CS Department
Jump to navigation Jump to search
← Older editNewer edit →
Charliep (talk | contribs)
Bureaucrats, nsAppliedGroupsD, nsAppliedGroupsR, nsAppliedGroups_, nsCS382DEL, nsCS382RW, nsCS382_TalkDEL, nsCS382_TalkRW, nsDepartmentDEL, nsDepartmentRW, nsDepartment_Tal, nsFundingDEL, nsFundingRW, nsFunding_TalkDE, nsFunding_TalkRW, nsGradingDEL, nsGradingRW, nsGrading_TalkDE, nsGrading_TalkRW, Administrators
1,979 edits
mNo edit summary
Charliep (talk | contribs)
Bureaucrats, nsAppliedGroupsD, nsAppliedGroupsR, nsAppliedGroups_, nsCS382DEL, nsCS382RW, nsCS382_TalkDEL, nsCS382_TalkRW, nsDepartmentDEL, nsDepartmentRW, nsDepartment_Tal, nsFundingDEL, nsFundingRW, nsFunding_TalkDE, nsFunding_TalkRW, nsGradingDEL, nsGradingRW, nsGrading_TalkDE, nsGrading_TalkRW, Administrators
1,979 edits
No edit summary
Line 1: Line 1:
== Big-FE - A cost and power efficient platform for high performance computing education ==
== Big-FE - A cost and power efficient platform for high performance computing education ==


Alex Lemann(*), Josh McCoy(*), Charles Peck.  Cluster Computing Group, Earlham College.
Alex Lemann(*), Josh McCoy(*), Tobias McNulty(*), Charles Peck.  Cluster Computing Group, Earlham College.


Abstract:
Abstract:


There is a shortage of high performance computing resources for educational use.  Big-FE, short for Focus on Education, is a complete design and hardware prototype for a cost effective, electrically efficient, scalable high performance computing resource.  Our design is a synergy of the classic Beowulf commodity-off-the-shelf cluster philosophy and the commercial, proprietary blade approach to computational cluster design.  Big-FE leverages the strengths of each approach to provide a cost effective computing platform for computational science education.   
There is a shortage of high performance computing resources for educational use.  Big-FE, short for Focus on Education, is a complete design and hardware prototype for a cost effective, energy efficient, scalable high performance computing resource.  Our design is a synergy of the classic Beowulf commodity-off-the-shelf cluster philosophy and the commercial, proprietary blade approach to computational cluster design.  Big-FE leverages the strengths of each approach to provide a cost effective computing platform for computational science education.  Big-FE uses a diskless compute node architecture running a Liberated Bootable Cluster CD single system image.   


We measure and compare the electrical consumption of the Big-FE prototype with popular commercial high performance computing platforms.  We describe our physical layout which is organized to minimize both maintenance and cooling costs.  We detail the cost structure, both up-front and on-going, and estimated performance for Big-FE and compare these with popular commercial high performance computing platforms.   
We measure the energy consumption of the Big-FE prototype and compare it with popular commercial high performance computing platforms.  Our physical layout is organized to minimize both maintenance and cooling costs.  We detail the cost structure, both up-front and on-going, and estimated performance for Big-FE and compare these with popular commercial high performance computing platforms.   


===Notes===
===Notes===

Revision as of 19:14, 12 May 2006

Big-FE - A cost and power efficient platform for high performance computing education

Alex Lemann(*), Josh McCoy(*), Tobias McNulty(*), Charles Peck. Cluster Computing Group, Earlham College.

Abstract:

There is a shortage of high performance computing resources for educational use. Big-FE, short for Focus on Education, is a complete design and hardware prototype for a cost effective, energy efficient, scalable high performance computing resource. Our design is a synergy of the classic Beowulf commodity-off-the-shelf cluster philosophy and the commercial, proprietary blade approach to computational cluster design. Big-FE leverages the strengths of each approach to provide a cost effective computing platform for computational science education. Big-FE uses a diskless compute node architecture running a Liberated Bootable Cluster CD single system image.

We measure the energy consumption of the Big-FE prototype and compare it with popular commercial high performance computing platforms. Our physical layout is organized to minimize both maintenance and cooling costs. We detail the cost structure, both up-front and on-going, and estimated performance for Big-FE and compare these with popular commercial high performance computing platforms.

Notes

  • Topics to include:
    • Power consumption
    • Packaging and density
    • Cost of purchasing
    • Cost of ownership
Retrieved from "https://wiki.cs.earlham.edu/index.php?title=Big-FE-teragrid-abstract&oldid=1684"