Hi there, We're currently attempting to explain a slowdown of an LVQ-type parallel analysis we're working on. We are benchmarking our analysis running over openMosix against the same running via SNOW for R. Both perform similarly on small datasets, but on large datasets SNOW drastically outperforms openMosix. However, these results are achieve running SNOW on the default RedHat kernel (Enterprise Edition, RHEL-3) and mosix on the same kernel patched with the rpm for RedHat 9 from sourceforge. Even though the rpm seems to be compatible and everything runs fine, when the same SNOW analysis is run under the patched kernel there is an enormous slowdown in runtime, to the point that openMosix outperforms SNOW. Is there some sort of incompatibility issue with the RedHat kernels (either general or specific to this version) that would cause a slowdown of this kind? I know that Enterprise Edition was marketed as having tools to increase java clustering speeds, would this interfere with the type of socket communication SNOW uses? Thanks, Jim
Samuelson, Frank*
2004-May-03 19:51 UTC
[R] openMosix vs SNOW: redhat kernel causing slowdown?
You're using sockets for connection in snow? or pvm or mpi? There's nothing magical about snow. It just uses the socket connections provided in R, which in turn uses regular BSD sockets. -----Original Message----- From: Jim Thomas [mailto:james at staarfunds.com] Sent: Thursday, April 29, 2004 6:13 PM To: r-help at stat.math.ethz.ch Subject: [R] openMosix vs SNOW: redhat kernel causing slowdown? Hi there, We're currently attempting to explain a slowdown of an LVQ-type parallel analysis we're working on. We are benchmarking our analysis running over openMosix against the same running via SNOW for R. Both perform similarly on small datasets, but on large datasets SNOW drastically outperforms openMosix. However, these results are achieve running SNOW on the default RedHat kernel (Enterprise Edition, RHEL-3) and mosix on the same kernel patched with the rpm for RedHat 9 from sourceforge. Even though the rpm seems to be compatible and everything runs fine, when the same SNOW analysis is run under the patched kernel there is an enormous slowdown in runtime, to the point that openMosix outperforms SNOW. Is there some sort of incompatibility issue with the RedHat kernels (either general or specific to this version) that would cause a slowdown of this kind? I know that Enterprise Edition was marketed as having tools to increase java clustering speeds, would this interfere with the type of socket communication SNOW uses? Thanks, Jim ______________________________________________ R-help at stat.math.ethz.ch mailing list https://www.stat.math.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide! http://www.R-project.org/posting-guide.html