—Quick and incremental growth in the processor power of desktop personal computers and network bandwidth due to recent extraordinary technological advances, have shifted the trend of parallel processing from conventional costly massively parallel supercomputers to the comparatively inexpensive cluster of networked desktop PCs for solving data and computation intensive sequential as well as parallel applications. For such parallel applications, cluster of LAN based networked PCs environment has become the boon in developing countries because of easy availability of relatively inexpensive computational resources. This paper presents a parallel computing framework based on cluster of networked desktop PCs that intends to optimally exploit the pooled computational strength of networked desktop PCs available in the intranet of university campus. This Cluster Based Parallel Computing framework (CBPCF) is based on the Master-Slave computing paradigm and it emulates the parallel computing environment. Performance statistics of such a cluster based framework is evaluated using experimental setup by running applications like parallel Matrix multiplication and Pi(Π)value approximation. Interpretation of results has shown that high bandwidth requirements in problems like matrix multiplication ,is a major hindrance to get good performance as major percentage of the turnaround time is consumed as communication time. In Contrary to matrix multiplication application, Pi approximation problem has shown good amount of speedup as well as efficiency due to more computation work involved than communication in the problem.
—Cluster, CBPCF, High Performance Computing, Master-Slave Computing paradigm
Amit Chhabra is with department of Computer Science & Engineering, Guru Nanak Dev University, Amritsar, India.
Gurvinder Singh is with department of Computer Science & Engineering, Guru Nanak Dev University, Amritsar, India.
Cite: Amit Chhabra, Gurvinder Singh, "A Cluster Based Parallel Computing Framework (CBPCF) for Performance Evaluation of Parallel Applications," International Journal of Computer Theory and Engineering
vol. 2, no. 2, pp. 226-232, 2010.