Level of Education of Students Involved

Undergraduate

Faculty Sponsor

Nicholas Rosasco

College

Arts and Sciences

Discipline(s)

Computer Science

ORCID Identifier(s)

Christian Garcia: 0000-0002-0591-4368 Nicholas Rosasco: 0000-0001-9726-9986

Presentation Type

Poster Presentation

Symposium Date

Spring 4-25-2024

Abstract

Parallel and distributed computing (PDC) and high performance computing (HPC) tools and techniques are becoming increasingly common and even necessary in many disciplines, particularly as large data sets and high computational loads become commonly encountered challenges. Experimentation in this space is somewhat constrained by the cost and availability of the assets that handle even small and midsize tasks in this space; initial tests and bench-marking may be the only chance to put systems through their paces. Therefore, it has become important to provide information on optimizing these smaller but still impactful systems. This requires experiments focused on delivering more capability with commodity options and configurations.

A series of HPC tools were identified for testing against commonly used benchmarks that use parallelization techniques such as MPI and OpenMP. For the selection of each benchmark, networking speed was considered above CPU power as this may be a road to increased performance at minimal additional cost. The configuration used for this experiment consists of a computing cluster with three AMD Ryzen 9 5950 CPUs and one AMD Ryzen 9 7950X CPU, all readily available. Variations in performance were investigated using the selected benchmarks: specific tests were done with both homogeneous and heterogeneous CPU configurations, as well bonded and unique-IP configurations for 10GbE network adapters.

Biographical Information about Author(s)

Christian Garcia is an undergraduate student graduating in 2024 with a B.S in computer science and a minor in mathematics.

Download

Share

COinS