Integrating Linpack Benchmark with Condor for Optimal HPC Cluster Performance
Hey everyone @Middleware & OS , I'm currently benchmarking a small HPC cluster (4 nodes, each with an Intel Xeon E5-2630 CPU) running CentOS 7.9. The cluster uses Condor (v8.9.3) with MPICH2 for job scheduling and communication. I've installed Linpack and the ATLAS library, aiming to achieve the maximum GFLOPS for the cluster.
I've been looking at the Condor documentation and online resources, but I haven't found a clear guide on integrating Linpack with Condor for benchmarking. Specifically, I'm not sure how to submit Linpack jobs as Condor jobs, how to configure Condor to distribute the workload better across the cluster nodes using MPICH2.
Pls I need advice on achieving this.
Solution:Jump to solution
@Dtynin
I would have suggested you use debian instead of centOS, but you can Ensure Linpack and MPICH2 is correctly installed and functioning on the cluster
create a bash script to collect input run executables, display the outcome and monitor execution and performance results to evaluate the maximum GFLOPS achieved by the cluster...
14 Replies
Why centOS?
CentOS is a popular choice for HPC clusters due to its stability, long-term support, and compatibility with many scientific and engineering applications. it also has a robust community and extensive documentation, making it easier to find support and resources for managing and optimizing HPC workloads, and can be used also to build embedded systems
It’s EOL’ing next month so wondering if you’re thinking about migrating to another distro
Try out another distro to work conveniently with
Our server was also built in with centos, @Dtynin sorry i can't help cause this is not my field,but centos this the best distro four this kind of use, good Luck
Are you all going to migrate?
It was my previous company but i don't think they will, since it is working for years, they keep updating the os, and like we say in french (on ne change pas une équipe qui gagne) Never change a winning team,
it all depends on application , CentOS has slow updates
Well maybe they prioritize on stability
Solution
@Dtynin
I would have suggested you use debian instead of centOS, but you can Ensure Linpack and MPICH2 is correctly installed and functioning on the cluster
create a bash script to collect input run executables, display the outcome and monitor execution and performance results to evaluate the maximum GFLOPS achieved by the cluster
I agree with you
it's all good, thanks
I would try to consider that
this is quite helpful, thanks @Enthernet Code