出版社：SISSA, Scuola Internazionale Superiore di Studi Avanzati
摘要：The floating point computation runs at the Central Information and Computing Complex (CICC)
cluster of JINR accommodate requests for traditional sequential applications, parallel computing
applications, as well as Grid applications launched within various virtual organizations. During
the second half of 2008, the CICC JINR comprised a heterogeneous computing array of multicore
processors with a total of 560 cores grouped in seven modules. Performance measurements made
in 2007 and system exploitation evidenced the existence of bottlenecks which were identified and
alleviated. As a result of the implemented optimizations, the system works efficiently for all three
abovementioned categories of jobs. Within the Russian Data Intensive Grid (RDIG) consortium,
which comprises, besides the CICC JINR, 14 Russian computing centres, our cluster covered a
sizeable part of the RDIG share to the LHC projects during 2007–2008. Performance measurements
using the High Performance LINPACK Benchmark showed relative figures at the level of
the best results reported in the June 2008 TOP500 edition of the most performing computers in
the world.
