Research Of Data Storage And Query Processing In Wireless Sensor Networks

A wireless sensor network is composed of a large number of sensor nodes deployedin the monitored area. These nodes collaborate in an ad hoc way to sense, sample andprocess information obtained from objects monitored by the network. As sensors con-stantly sample data from the environment, a sensor network can be seen as a specialkind of database. Users acquire useful information by querying the network. Sensor net-works have many features, such as a majority of nodes, short communication range andlimited resources. The most important feature of a sensor network is each node has lim-ited energy. Query processing algorithms for sensor networks must reduce their energyconsumption as much as possible. Data storage method is an important part of a queryprocessing algorithm. Organizing data produced by sensors in an efficient way can savethe energy consumption of a query processing algorithm and enlarge the working timeof the sensor network. In this paper, we study the problem of data storage and relatedquery processing algorithms in the sensor network. Main contributions of our work areas follows:A time index based data storage method and a corresponding query processing al-gorithm are proposed to solve the energy waste of the traditional data centric storagemethod during the procedure of the event storage for a source node frequently detectingevents in some particular place of a sensor network. In this case, it is more efficient tosave data in a source node itself and the source node stores an index on the data stor-age node to indicate where the source node is. Then a user's query can find results onthe source node according to the index. We first analyze the condition in which the in-dex based data storage algorithm can save more energy than the traditional data centricstorage algorithm and present an adaptive data storage selection algorithm according tothe condition. Compared with the traditional data centric storage algorithm, the adaptivealgorithm can dynamically adjust the data storage method in real time according to therelationship between the number of events and queries. Experimental results show thatthe adaptive data storage selection algorithm can save more energy than the traditionaldata centric storage algorithm.A ring based load balance data storage method and a query processing algorithmbased on such kind of data storage method are proposed to deal with the hotspot problemfor data centric storage method when events are evenly distributed in a sensor network. Ontheonehand,theringbasedloadbalancedatastoragemethoddividesasensornetworkinto many rings and stores data among all nodes within a ring. At the same time, differentringsalternativelyworkduringtheworkingtimeofthenetwork. Bothofthetwomethodsavoid the hotspot problem during the procedure of data storage. On the other hand, thequery processing algorithm based on the load balance data storage method makes allnodeswithinaringcooperativelyanswertheuser'squery. Bymakingallnodeswithinthenetwork consume energy evenly, our algorithm extends the lifetime of a sensor network.Experimentalresultsshowthatourringbasedloadbalancedatastoragemethodandqueryprocessing algorithm can solve the hotspot problem and prolong the lifetime of a sensornetwork.A Filter storage based query processing algorithm is proposed to solve a new kindof continuous query named continuous dynamic range query. The difference betweenan ordinary range query and a dynamic range query is the range of the former one isstatic. The range of an ordinary range query does not change during the execution timeof a query, while the range of our dynamic range query can constantly change during theexecution time of a query. In summary, the meaning of the dynamic range query is to usethe real time reading of some sensor in a sensor network as the comparison target. Theuser wants to get readings of other sensors, which are larger than the target's reading ineach period. We proposed a Filter storage based query processing algorithm to solve theproblem. Each sensor is assigned a Filter based on which a sensor decides whether tostore the reading locally or externally. The critical problem is to select a proper Filter foreach sensor to minimize the energy consumption of a query and we propose an energybased algorithm to solve this problem. The Filter storage based algorithm can reduce theflooding times of the local storage based algorithm, meanwhile it avoids irrelevant nodestransmitting data to sink for the external storage based algorithm. Experimental resultsshow that the Filter storage based algorithm can save more energy than other algorithms.A local storage based energy efficient query processing algorithm is proposed tosolve a new kind of snapshot query, named the node number constraint query for a sen-sor network. The meaning of the node number constraint query is only a part of nodes,satisfying a user's query, return their results to sink and the number of nodes returningtheir results must satisfy the user's precision request. Compared with other algorithmsadopting local storage method, our algorithm fully utilizes the feature of the node numberconstraint query to select relevant nodes, which can not only reduce the energy waste offloodingquerythroughoutthenetworkbutalsosavetheenergyconsumptionofresultcol-lection. The query processing algorithm is composed of three parts: relevant node selec- tion algorithm, energy efficient query dissemination algorithm and energy efficient resultcollection algorithm. The relevant node selection algorithm selects a part of nodes in thenetwork as targets to send query at the same time guarantees the precision of the result. ASteiner tree based query dissemination algorithm are proposed to transmit a query to tar-gets in an energy efficient way. The energy efficient result collection algorithm presentstwo kinds of result collection strategies, named direct and indirect collection strategy,then the collection algorithm gives the condition, in which each strategy is suitable tocollect results. Experimental results show that the local storage based query processingalgorithm for node number constraint query can not only guarantee the precision of theresult but also save more energy than other algorithms.