The Research Of Middleware Architecture Technology In Wireless Sensor Network Based On Mapreduce

With the rapid development of information industry, the wireless sensor network(WSN) is becoming a hot research point of the Internet of things. The integration ofsensor, network and micro motor system make WSN has a broad prospect ofapplication in the areas of medical, industrial and agriculture.Middleware technology in wireless sensor network is an important part of thedesigning of WSN. It can provide the services and data application to the uppermonitor which is needed by shielding the Heterogeneity of underlying hardware. Andthe application of middleware technology are concentrated in specific area, such asenergy management, data processing etc. As the extensive of the Internet of things time,the requirements of a general middleware framework with lightweight, which issuitable for mass data process in sensor network has become increasingly urgent. TheMapReduce technology introduced by Google is a data process model based on thecloud computing, with good ability of data processing. And the topological structure ofprocessing mechanism in each layer, both in cloud computing and wireless sensornetwork is very similar.Based on the analysis and classify of the traditional middleware technology inwireless sensor network, this paper perceives the program restructuring of Sensing andComputing by Map and Reduce technology, and proposes a framework of middlewarein Wireless sensor network based on MapReduce technology which firstly introducesthese two concept,Map and Reduce, into the wireless sensor network. This frameworkadopts the structure with three layers: the CPE layer released the user compiled code tothe nodes by the parser and the distributor, and display the results to achieveinteraction with user; the integrated layer realizes the task of Map and Reduce,includes all the nodes and interfaces; the Operating environment layer deployed onevery node, each node runs a byte code interpreter. Based on the three layers structure,this framework can support different operating systems of sensor network by changingthe code interpreter.In order to match the architecture based on the MapReduce, this paper designed acommand language to provide the programming abstraction model of wireless sensornetwork. The two program model of task descriptors and work descriptor arerespectively corresponding to Map and Reduce process, to realize the task predefiningand data gathering. The programming scheme can shield the diversity of underlyinghardware, thus to provide a programming approach which understand easily, and achieving data operation and data filtering of nodes.The application of MapReduce middleware which proposed in this paper onenergy management show that, this method can improve the traditional method ofcollecting energy by setting the energy threshold to comparison operation using thenode self-computing capacity. When the node energy lower than threshold, the nodecan immediately return the node ID and the energy data values <node_ID, battery>, sothat make the upper monitor to find the nodes with low energy, and take effectivemeasures, such as the replacement of the battery or battery charging. That ensures theWSN will not take place the large problems such as data acquisition with unstablewhich took place by the insufficiency function of nodes.The experiment results show shat, the WSN middleware framework based on theMapReduce proposed in this paper has excellent availability, good double operationand data dependence, at the same time, with the advantages of strong expression abilityand of regardless of the underlying hardware. As a result, this frame can lead theprogrammers easily achieve the whole programming of network, and has a goodperformance in the application of multi environment.