Distributed Parallel Processing Based on Master/Worker Model in Heterogeneous Computing Environment
Lei Wang, Yongzhong Huang, Guo Shao-zhong · INTERNATIONAL JOURNAL ON Advances in Information Sciences and Service Sciences · 2012
Due to the complexity and varying requirements of the applications to utilize large-scale computing resources, there are several issues such as aggregating heterogeneous computing resources, easy-touse programming model and fault-tolerant mechanism that need to be addressed. This paper presents a general distributed parallel processing architecture based on master/worker model, and it can aggregate various heterogeneous computing resources without coupling to any specific hardware and architecture. On this basis, we implement a general framework based on TSpaces for large-scale distributed computation of problems that feature coarse-grained parallelization. Furthermore, we present a dynamic managing mechanism of voluntary computing resources in unreliable environment. The experiment results indicate that our implementation has preferable performance, and can achieve high availability and reliability in the environment of dynamic changing resources and unreliable loose-coupling networks.