Cluster Analysis

Project Details

  • Role : M.Sc Thesis - High Integrity Systems (Computer Science and Engineering)

  • Details : Running parallel programs on a traditional cluster requires expensive and special processing resources to obtain the required result within a time period. Currently, in the world of computers, a single-board computers are very popular and efficient because of the low power consumption, inexpensive and tiny size. This thesis work shows how well the ODROID-U3, a singleboard computer performs in the cluster environment

    The constructed cluster is based on the 4 ODROID-U3 nodes. NFS is used as a clustered file system which is shared across the computing nodes of the cluster. One of the popular standards of MPI, the MPICH2 library used in cluster to run parallel applications. In order to perform basic vector and matrix operation BLAS library is used. For the performance analysis of MPI program on ODROID cluster, three metrics named as execution time, efficiency and speedup are used.

    This thesis describes the construction of Linux cluster based on ODROID-U3, single board computer and performance analysis of it. In order to evaluate the real-time performance, acceptability and usage of the cluster, various types of benchmark test suits and some of the parallel programs are executed.