ANALISIS KINERJA HIGH PERFORMANCE COMPUTING PADA KOMPUTASI PARALEL UNTUK MENYELESAIKAN INTEGRAL NUMERIS

• Main Authors: , Muhammad Aditya Pradana, , Teguh Bharata Adji, S.T., M.T., M.Eng., Ph.D
• Format: Theses and Dissertations NonPeerReviewed
• Published: [Yogyakarta] : Universitas Gadjah Mada 2011
• Subjects: ETD
• Online Access: https://repository.ugm.ac.id/91021/; http://etd.ugm.ac.id/index.php?mod=penelitian_detail&sub=PenelitianDetail&act=view&typ=html&buku_id=52860
• Institution: Universitas Gadjah Mada

Along with the increasing need to solve various problems of numerical computation effectively and efficiently, the need for a computer system with high computing capability has increased. Computer systems with high computing capability offers the ability to integrate the resources of multiple computers to solve a problem of numerical computing. This computer system called computer cluster. The cluster must have the ability to perform computing process by using parallel computing mechanism called message passing. In this study, the implementation of message passing mechanism on a computer cluster is done by using Open Message passing Interface (OpenMPI) application. This study aims to analyze the performance of computer cluster using MPI mechanism in handling the process in parallel computing based on the execution time, speedup, and the efficiency. Parallel computing processes will be executed with OpenMPI application to solve the problems of numerical integration using trapezoidal method.The research method used in this study is by implementing OpenMPI on a Linux-based cluster systems and then analyze the performance of the system in dealing with parallel computing process to solve the numerical integration problems by increasing the number of intervals used on a numerical integration problem. small number of integration intervals The results of this study showed that cause the parallel execution time slower than sequential execution time, although number of nodes increment done in the parallel execution. With the use of high number of integration intervals, parallel execution is faster as the number of nodes increase. By increasing number of intervals, the sequential execution time is initially faster than parallel execution time, although finally reduced significantly with the increase of number of integration interval. With limited number of nodes, the use of send and receive routine produces a significantly faster execution time than broadcast and reduce routine that known as dynamic message passing communications model. Keyword : high performance computing, cluster, message passing, parallel computing, numerical integration, OpenMPI, execution time, speedup, efficiency