A Study Optimal Model of Task Partitioning in a Heterogeneous Parallel Computing

Efficiency and Flexibility in Task Partitioning for Heterogeneous Parallel Computing

Abstract

Optimal Task Partitioning with Duplication (OTPSD) restricts schedule duration just as the overhead correspondence. Duplication of functions has reduced the overhead coordination of the computational system. It is three step algorithms where the first stage requires the design of grain pack Sub DAG. The following stage is a task process of need. The processors are grouped in the third stage by their processing capacities. The proposed algorithm (OTPSD) constraints improve flexibility and display efficiency over MCP and HEFT algorithms over uniform calendar length and processor usage. Determined compute unified system architecture (CUDA) rendering length of OTPSD is (184.4 Sec) shorter than MCP (215.6 Sec) and HEFT (196.3 Sec) algorithms.