Benchmarking of High Performance Cluster Reynolds
Main Article Content
Abstract
This paper demonstrates a complete benchmarking exercise of a high performance cluster Reynolds at Tata Steel R&D division. A complete bench script has been reported for running parallel programs in queue and that can be directly used with all parallel computations purposes. Randomness in node distribution is found in batch processes. The scalability of the computations on a 80-processor SUSE Linux cluster is evaluated for smaller to large size problems. It is found that for a case with 0.417 million cells, after 32 processors performance curve flattens and then reduces marginally. This is due of partitioning the case into too many cores, which increases the communication between the cores and the scalability does not remain linear. For all other cases a super-liner speed up in computational efficiency and system functional rating is observed. System scalability study is performed for a live industrial case with OpenFOAM. For such case after 40 numbers of processors, reduction in speedup is observed, whereas there is a sharp reduction in clock time till 80 processors. Reynolds performance over the available supercomputing benchmarks is compared. A new parameter Reynolds performance index is defined to compare cluster’s performance over other high performance machines. Depending upon the problem complexities and mesh sizes, recommendations have been given to limit the optimum node usage. This is for the first time such exhaustive study in benchmarking exercise being reported
Keywords:Reynolds, HPC, benchmarking, rating, efficiency
Downloads
Article Details
COPYRIGHT
Submission of a manuscript implies: that the work described has not been published before, that it is not under consideration for publication elsewhere; that if and when the manuscript is accepted for publication, the authors agree to automatic transfer of the copyright to the publisher.
Authors who publish with this journal agree to the following terms:
- Authors retain copyright and grant the journal right of first publication with the work simultaneously licensed under a Creative Commons Attribution License that allows others to share the work with an acknowledgment of the work's authorship and initial publication in this journal.
- Authors are able to enter into separate, additional contractual arrangements for the non-exclusive distribution of the journal's published version of the work (e.g., post it to an institutional repository or publish it in a book), with an acknowledgment of its initial publication in this journal.
- Authors are permitted and encouraged to post their work online (e.g., in institutional repositories or on their website) prior to and during the submission process, as it can lead to productive exchanges, as well as earlier and greater citation of published work
- The journal allows the author(s) to retain publishing rights without restrictions.
- The journal allows the author(s) to hold the copyright without restrictions.