Research On The Synergism Between Distributed Storage And Parallel Computing In Cloud Computing Environment

Cloud computing is one of the distributed computing environment witch is built on thehomogeneous cheap ordinary nodes. Because of “nodes are not credible”, distributed storagebecomes the inevitable choice of cloud computing, and the integration trend of storage andcalculation is becoming obviously. This paper built a simple peer-to-peer cloud computingenvironment, and studied both aspects of the possible collaboration: distributed storage andparallel computing, finally put forward a set of distributed storage strategy and a data-drivenoriented scheduling policy.The physical layer of the cloud computing environment in this paper is built on in thecabinet, and the network topology is peer-to-peer. The environment is based on Springframework and apply some Java embedded components and popular Web frameworktechnology and communications architecture, which can realize rapid development, highlyscalability and rapid deployment. In this paper, the traditional MPI environment and thedistributed storage are integrated together, which make parallel computing without usingMap/Reduce to be possible.In the aspect of distributed storage, this paper puts forward a set of distributed storagestrategy which is suitable for seismic data in cloud computing environment, which involvesdata partitioning, replica deployment and positioning based on full study of the seismic dataand its processing characteristics. In the aspect of parallel computing, this paper also putforward a data-driven oriented peer to peer scheduling strategy except for the combination ofthe distributed storage and MPI. The strategy not only amortize the cost of distributed storage,but also save the time of data preparation by making full use of the advantages of distributedstorage.