Hi James: As you have guessed, "benchmark" really varies depending on what you are trying to do. The best benchmark is usually the application you will run the most on your cluster - if you can't get your application to run favourably on your cluster, then it doesn't really matter what linpack tells you your cluster can compute.
3DMark is usually a benchmark used on PCs to determine how well its 3D performance is, and is not usually used for benchmarking HPC clusters. However, if you are into GPU clusters, then perhaps that's the benchmark for you. I'd like to take this opportunity to plug Joe Landman's BBS, although the name implies it's for "Bioinformatics" benchmarking, it can in fact plug in different software packages outside the field of Bioinformatics (eg. CFD, oil & gas, etc.): http://www.scalableinformatics.com/metadot/index.pl?id=2194 Cheers, Bernard -----Original Message----- From: [EMAIL PROTECTED] on behalf of Otter Sent: Sat 16/09/2006 12:17 To: [email protected] Subject: [Oscar-users] Cluster Benchmarks Greetings, I am fairly new to cluster computing. I have constructed a cluster and would like to benchmark it. I know the term benchmark is broad, but I do not know if there are standards for cluster computing such as PCMark, 3DMark etc. I am working with a group, so if my question seems to lack details it is because I do not know all of the details. At the moment I just know we will need to generate some benchmarks as we experiment with the cluster. I noticed a link to "A Portable Implementation of the High-Performance Linpack Benchmark for Distributed-Memory Computers" on the OSCAR website ( http://www.netlib.org/benchmark/hpl/ <https://webmail.latech.edu/horde/services/go.php?url=http%3A%2F%2Fwww.netlib.org%2Fbenchmark%2Fhpl%2F> ), but before I blindly selected a package I would like to ask for input from real cluster users. Are there certain statistics that matter most? Is there a preferred benchmarking package? Is there a preferred way to benchmark a cluster? I do understand that certain statistics are more meaningful based on what type of cluster computing you are doing, but I would rather gather to many statistics rather then to few. And I want my results to be as professional as possible. Thank you for any input, James ------------------------------------------------------------------------- Using Tomcat but need to do more? Need to support web services, security? Get stuff done quickly with pre-integrated technology to make your job easier Download IBM WebSphere Application Server v.1.0.1 based on Apache Geronimo http://sel.as-us.falkag.net/sel?cmd=lnk&kid=120709&bid=263057&dat=121642 _______________________________________________ Oscar-users mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/oscar-users
<<winmail.dat>>
------------------------------------------------------------------------- Using Tomcat but need to do more? Need to support web services, security? Get stuff done quickly with pre-integrated technology to make your job easier Download IBM WebSphere Application Server v.1.0.1 based on Apache Geronimo http://sel.as-us.falkag.net/sel?cmd=lnk&kid=120709&bid=263057&dat=121642
_______________________________________________ Oscar-users mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/oscar-users
