Home

About

Advanced Search

Browse by Discipline

Scientific Societies

E-print Alerts

Add E-prints

E-print Network
FAQHELPSITE MAPCONTACT US


  Advanced Search  

 
Computational physics Superlinear speedup on NANCO -the new RBNI cluster computer
 

Summary: Computational physics
Superlinear speedup on NANCO - the new RBNI cluster computer
Pavel Bavli1, and Joan Adler1,
1Physics Department, Technion
There have been many arguments over the value of centralised high performance com-
puters (HPC) over local workstations. Israel came late to HPC because of export license
politics and our original public facilities became outdated quickly. The revival of HPC in
the rest of the world during recent years was also slow to percolate here.
However, after many years of no local academic access to a high performance com-
puter, RBNI at the Technion purchased NANCO for computations in nanotechnology in
late 2006. It is a 128 dual-core processor (2.2 GHz AMD Opteron) LINUX cluster with a fast
VOLTAIRE interconnect based on DDR Infiniband, provided by SUN/EMET. It has been
fully operational since mid-summer, with a good range of compilers and MPI for paral-
lelization. Details about NANCO can be found at
http://phycomp.technion.ac.il/ nanco
One of the first projects to be completed on NANCO is a study of its speedup statistics for
a molecular dynamics simulation of the elastic constants of argon. Up to a million particles
have been modeled (using only 6.25 % of the machine) but the extensive statistics are for
1/4 million, equivalent to a 21.6nm sample. Using MPI with spatial decomposition runs of
2500ps took 177,330 seconds for 4 processors and only 76,366 for 8. For 16(20) processors

  

Source: Adler, Joan - Physics Department, Technion, Israel Institute of Technology

 

Collections: Physics