Efficient and Robust Query Processing for Mobile Wireless Sensor Networks

We present CNFS, an algorithm for efficient and robust query processing for mobile wireless sensor networks. CNFS is a walk-based algorithm that is biased to visit nodes close to the source first. This bias is accomplished by collec ting topology information about the network as the search progresses. This information is also used to tolerate changes in the network topology caused by node mobility that could otherwise cause the query to fail. As a result, CNFS requires fewer messages to process a query than flooding-based algorithms, while toler ating node mobility better than random walk-based algorithms. Our experiments show that in medium-density networks (average node degree 8.3) CNFS requires about 37% fewer messages than the other algorithms studied, while experiencing significa ntly fewer query failures than random walk-based algorithms in both sparse and dense networks. CNFS's success rate is comparable to flooding-based algorithms in dense networks and slightly wo rse in sparse networks. I. INTRODUCTION We present an efficient and robust algorithm, called CNFS (Closest Neighbor First Search) for query processing in mobile wireless sensor networks. To process a query a base station initiates a query message that is propagated among the sensors until the query is satisfied. The response is returned to the b ase station along a path discovered during the query propagation. Two of the most important metrics when evaluating query processing schemes are efficiency and robustness. Efficienc y is represented as the number of messages required to satisfy the query; this is important because sensor nodes normally have very limited power and message transmissions dominate a node's power consumption. Excessive messages can also increase network congestion. Robustness is also important because queries may fail due to mobility and unreliable com- munication. An acceptable query processing scheme should maximize the success rate despite network dynamics. Com- pared to algorithms based on flooding or random walk, our algorithm requires fewer message transmissions and is robust against changing network topologies. Conventional query processing algorithms use blind search to satisfy the query - that is, no global information about th e network is gathered prior to initiating the search. Such inf or- mation might include a hierarchy of nodes or an index of the data they contain. This information would obviously be useful, but is impractical to maintain in the face of mobile nodes that cause the network topology to change rapidly. The potentially high data generation rate from sensors in a sensor network also makes the maintainance of the indexing mechanism costly. For these reasons query processing algorithms typically employ either flooding or random walk. Flooding floods the network with the query to find the answer. It returns the answer very quickly, and is therefore highly-tolerant of changing network dynamics, but it requires an excessive number of messages and can congest the network. In contrast, random walk uses a single message that randomly visits the network nodes. This reduces the message overhead and avoids congestion, but can be very slow and may not complete at all if the network topology changes such that the source or destination cannot be visited. For this reason it has a much lower success rate than flooding. CNFS is based on a directed blind search, so that no pre-existing structure is imposed on the nodes or the data they contain. The search is directed by topology information collected as it progresses, allowing CNFS to be both efficien t and robust. CNFS collects and maintains a partial map of the network during query processing, from which it finds an optimal order of visiting nodes. Our algorithm uses a biased search scheme, i.e., whenever there are several unvisited nodes to choose from as next step, it chooses the one closest to the source node. The partial map is also used to compute the shortest return to the source, as well as determine alternat ive routes when links break, helping to improve both efficiency and robustness. For example, our experimental results show that for query processing in a 100-node network with average node degree of 8.3, CNFS requires about 37% fewer messages compared with the other methods. In addition, its success rate is significantly higher than random walk-based algorithms f or all network densities, and comparable to flooding in dense networks. The rest of the paper is organized as follows: Section II contains related work; Section III provides a formal descri p- tion of the problem; Section IV provides the details of the CNFS algorithm; Section V contains experimental results; and Section VI concludes.