Adopting graph reduction to synthesize parallel computation models

The depth of analysis and processing for big data has become an important factor of smart city construction. The existing data parallel computing models present the characteristics of diversification, pertinence and short cycle, which increase the development and maintenance difficulty of the models, impede the model standards to form. There lack an efficient method to synthesize different parallel computing models. In this paper we reduce the workflow graphs of parallel computing models by graph reduction of functional language after investigating the synthesized methods of parallel computing models, and then according to the reduced workflow graph to synthesize the kinds of models. Meanwhile we adopt resource tree expression and the method of reducing resource tree by graph reduction to describe and manage resources. This synthesized model not only has the high performance modules inherent in every one-model, it also has the dynamic reconfiguration function of the resources.