Parallel Computing Technology For Sensor Network Data Stream Based On Network Coding

With the development of IOT,the wireless sensor network which responsible for networking terminal information collection has been extensively applied in various fields.Cloud computing technology aims to process large scale data set professionally and effectively.The combined of wireless sensor network technology and cloud computing technology has turned out to be a new development trend.The system of parallel computing for wireless sensor network based on network encoding has combined with wireless sensor network technology and cloud computing technology,in order to achieve fast clustering of abnormal sensing data in large scale data set,to provide favorable guidance for users.The main work of this paper is as follows:(1)The wireless sensor network has low computing power and limited energy.This paper we used the Reed-Solomon code to devise a method of constructing a sparse matrix to improve the data transmission reliability and the energy utilization in wireless sensor networks.Through experiments testified that the method we designed can improve the energy utilization in the wireless sensor network.(2)Decoding will cause sink delay,and the computational will increase the burden of the decoding node.In this paper we design a distributed decoding algorithm for Hadoop cluster,we complete the decoding work in a cluster.This can improve the efficiency of decoding by using the powerful computing capability of the cluster,and reduce the burden of the node of a wireless sensor network.The feasibility of cluster parallel decoding is demonstrated by experiments,and the influence of the related factors on decoding efficiency is also demonstrated.(3)In view of the k-means algorithm under the framework of MapReduce will cause excessive consumption of I/O,this paper proposed an improved MapReduce algorithm based on single pass K-means,used for cluster analysis of sensing data.In this paper,we have proved that the proposed method can reduce the consumption of I/O theoretically when the program is executed.In the experiment,proved that the algorithm designed in this paper compared with the MapReduce k-means algorithm based on traditional clustering,in order to obtain the effect with less execution time.