MPI-Based Heterogeneous Cluster Construction Technology

Abstract

Nowadays, the applications of MPI-based clusters are used widely, such as the damage prediction of nuclear explosion and gene decryption. However, the usages of MPIbased heterogeneous cluster are nearly reported in contrast to the MPI-based homogeneous clusters due to the immature technology. This paper focuses on building a MPI based heterogeneous cluster with the MPICH2, which is released last year and has not supported the heterogeneous cluster yet. Beside, we give the reasons and test results in the paper to illustrate the benefits of using the MPI-based heterogeneous cluster without virtualization.

MapReduce Model Implementation on MPI Platform

“…However the MPI cannot give a satisfactory support to fault tolerance, which is the key requirement of the MapReduce model, so the earlier work of our team is focused on enhancing the MPI fault tolerant ability. A three-level fault tolerance mechanism has been implemented [8,11]:…”

3‐D data partitioning for 3‐level perfectly nested loops on heterogeneous distributed systems

“…Another approach is adding graphics processing units (GPUs) to improve performance of existing nodes . In the latter 3 cases, the cluster becomes heterogeneous . In this way, heterogeneous computing systems have emerged as an important contribution to provide computational capacity in high‐performance computing.…”

“…9 In the latter 3 cases, the cluster becomes heterogeneous. [10][11][12] In this way, heterogeneous computing systems have emerged as an important contribution to provide computational capacity in high-performance computing. In fact, the prevalence of heterogeneous systems in the TOP500 list grew from 3.4% to 18.0% between June 2010 and June 2015.…”