Red Storm/XT Supercomputers Cooling System Design and Optimization

Author

Yatskov, Alexander I.

Author_Institution

Cray Inc., Seattle, WA

fYear

2007

fDate

18-22 March 2007

Firstpage

30

Lastpage

36

Abstract

Developing the optimal design for a supercomputer system is a complex and time consuming process. This paper describes a sequential optimization approach that can lead to the most effective computer cooling system design in the shortest time possible. Ways of seeking effective Pareto designs are shown. Cost driven quality engineering was applied to develop cooling solutions for chips, printed circuit boards, cabinets and the computer facility. CFD thermal analysis, small and large scale experiments allowed to maximize supercomputer cooling system performance.

Keywords

Pareto optimisation; computational fluid dynamics; computer centres; cooling; mainframes; thermal analysis; thermal management (packaging); CFD thermal analysis; Pareto designs; XT supercomputers; computational fluid dynamics; cooling system; data center; heat sink; red storm supercomputers; Computational fluid dynamics; Cooling; Costs; Design optimization; Large-scale systems; Performance analysis; Printed circuits; Storms; Supercomputers; System analysis and design; Air Flow; Data Center; Design of Experiments; Heat Sink; Multicriteria Design; Optimization;

fLanguage

English

Publisher

ieee

Conference_Titel

Semiconductor Thermal Measurement and Management Symposium, 2007. SEMI-THERM 2007. Twenty Third Annual IEEE

Conference_Location

San Jose, CA

ISSN

1065-2221

Print_ISBN

1-4244-09589-4

Electronic_ISBN

1065-2221

Type

conf

DOI

10.1109/STHERM.2007.352389

Filename

4160883