Analysis of the CPU utilization for point-to-point communication operations in a Beowulf cluster system

Abstract

Beowulf cluster computing involves the use of a network of computing resources to provide high-performance computing (HPC) platform which possesses comparatively economical capabilities once reserved jbr supercomputers. Certain applications of this cluster computing should be looked from the problem solving viewpoint. Point-to-point communication between the cluster's nodes is fundamental to all communication subsystems as it forms as a basis for efficient message implementation, especially those that involve software applications. The network latency, the time required to transmit the message across the network, may generally be contributed by several significant factors. These may include the network bandwidth, underlying switching mechanism, and blocking lime. Based on the blocking time aspect, this parallel computing research attempts to analyze the CPU usage of the node by conducting a related benchmarking test.