Longjiang Guo

Power-Efficient node localization algorithm in wireless sensor networks

Sensor networks is an Ad-Hoc network consist of large amount of sensors. These sensors are distributed in a huge area. Each of the small, cheap, intelligent sensors has processor, memory and wireless transmission ability. These sensors collect or monitor the surroundings in real-time and process these data to obtain the detailed and accurate information from their covered area. Information from sensors must be combined with its location to make sense, so the location of sensors is very important in sensor network applications. Localization of sensors becomes one of the key techniques in sensor networks. A power-efficient localization algorithm is proposed in this paper. It uses few anchors(the sensors whose location is known) to implement the localization of other nodes without special devices such as GPS. This algorithm not only has the lower time cost and communication cost, but also needs only a few anchor nodes whose distribution is independent. Experimental results and analysis show that the localization algorithm proposed in this paper has higher accuracy, lower power cost and better expansibility. It is very suitable for large scale sensor networks.

Processing Sliding Window Join Aggregate in Continuous Queries over Data Streams

Processing continuous queries over unbounded streams require unbounded memory. A common solution to this issue is to restrict the range of continuous queries into a sliding window that contains the most recent data of data streams. Sliding window join aggregates are often-used queries in data stream applications. The processing method to date is to construct steaming binary operator tree and pipeline execute. This method consumes a great deal of memory in storing the sliding window join results, therefore it isn’t suitable for stream query processing. To handle this issue, we present a set of novel sliding window join aggregate operators and corresponding realized algorithms, which achieve memory-saving and efficient performance. Because the performances of proposed algorithms vary with the states of data streams, a scheduling strategy is also investigated to maximize the processing efficiency. The algorithms in this paper not only can process the complex sliding window join aggregate, but also can process the multi-way sliding window join aggregate.

Processing frequent items over distributed data streams

To improve the availability of communication bandwidth in distributed data stream systems, communication overhead should be reduced as much as possible under the constraint of the precision of queries. In this paper, a new approach is proposed to transfer data streams in distributed data stream systems. By transferring the estimated occurrence times of frequent items, instead of raw frequent items, communication overhead can be saved greatly. Meanwhile, in order to guarantee the precision of queries, the difference between the estimated and true occurrence times of each frequent item is also sent to the central stream processor. We present the algorithm of processing frequent items over distributed data streams and give the method of supporting aggregate queries over the preprocessed frequent items.

Dynamic Adjustment of Sliding Windows over Data Streams

The data stream systems provide sliding windows to preserve the arrival of recent streaming data in order to support continuous queries in real-time. In this paper, we consider the problem of adjusting the buffer size of sliding windows dynamically when the rate of streaming data changes or when queries start or end. Based on the status of available memory resource and the requirement of queries for memory, we propose the corresponding algorithms of adjustment with greedy method and dynamic programming method, which minimize the total error of queries or achieve low memory overhead. The analytical and experimental results show that our algorithms can be applied to the data stream systems efficiently.

Finding event occurrence regions in wireless sensor networks

Wireless sensor networks have emerged as a promising solution for a large number of monitoring applications. Sensor nodes are capable of measuring real world phenomena, storing, processing and transferring these measurements. However, users are interested in event monitored by sensors, but not the sensor itself or the massive irrelevant readings from sensors. Users often issue event queries such as “Where did happen hailstone in sensor network from 3:00 to 5:00?” Since battery supply of sensors is limited, energy-efficient query processing in sensor networks has become an important research problem. This paper presents an improved data-centric storage strategy, called CM-DCS, and also proposes two event query processing algorithms based on CM-DCS and local storage. The energy consumption of sensors for three storage strategies namely external storage, local storage and data-centric storage are analyzed and compared. The paper also studies the influence of the number of sensor nodes and node density on energy consumption. Analytical and experimental results show that in most cases the event query processing algorithm based on CM-DCS can save more energy than those algorithms based on external storage and local storage strategies.

An energy consumption estimation model for disseminating query in sensor networks

There are three approaches to disseminate queries into the sensor network, which are the multi-hop unicast, full flood and geocast. Since the energy is very limited resource of the sensor network, query optimizer must select an optimal approach consuming minimum energy to disseminate the query into the sensor network. As the selection relies on energy consumption estimation, the accuracy of the energy consumption estimation is very critical for query processing in sensor network. In this paper, a general energy consumption estimation model is proposed. Compared with other models, our model takes both routing protocol and MAC protocol into consideration. In experiments we compared our energy model’s estimation for the multi-hop unicast, full flood and geocast with simulation results. Results showed that our model’s accuracy is very high.

Evaluating Stream and Disk Join in Continuous Queries

In many stream applications, a kind of query that join the streams with the data stored in disk is often used, which is termed SDJoin query. To process SDJoin query, two novel evaluating algorithms based on buffer are proposed in this paper, namely BNLJ and BHJ. Since the existed cost models are not suitable for SDJoin evaluating algorithms, a one-run-basis cost model is also presented to analyze the expected performance of proposed algorithms. Theoretical analysis and experimental results show that BHJ are more efficient.