Original title:
Design and performance monitoring of a HPC cluster
Authors:
Jiroušek, Ondřej Document type: Papers Conference/Event: Doprava a telekomunikace pro 3. tisíciletí., Praha (CZ), 2003-05-26 / 2003-05-27
Year:
2003
Language:
eng Abstract:
For high performance computing (HPC), clusters of workstations are becoming very popular platforms, but performance of these systems is harder to predict than on the traditional massively parallel machines. The process by which a low-cost, high performance Beowulf-style Linux cluster was built is discussed in the paper. The first tests of several parallel tasks and the estimation of the performance of the cluster is shortly described.
Keywords:
linux cluster; message passing interface; parallel computing Project no.: CEZ:AV0Z2071913 (CEP), CEZ:MSM 210000024 Host item entry: Doprava a telekomunikace pro 3. tisíciletí, ISBN 80-01-02741-4
Institution: Institute of Theoretical and Applied Mechanics AS ČR
(web)
Document availability information: Fulltext is available at the institute of the Academy of Sciences. Original record: http://hdl.handle.net/11104/0106077