Sensor Net-based Online Analytical Processing

With the development of processor, memory and radio technology, networks composed of cheap nodes capable of sensing, communication and computation have become possible. The deployment of such networks can provide a large data source for the monitoring of physical world. Most existing systems for collecting and processing such sensor data only provide simple query operations, which may not be sufficient for users to get information that they want from the sensor data.In this paper, the on-line analytical processing (OLAP) operation, which has been widely used in traditional data warehouse system, is invited into the sensor network as a new way for users to query and analysis spatial related statistical result of environmental data such as temperature and humidity. The definition of the operation on sensor network called SOLAP is provided in this paper. To implement the operation on sensor network, many problems should be studied. First of all, the network should be organized properly to form a routing structure so that the data can be effectively sent to the user. In this paper the network is organized to a routing tree to accomplish it. An algorithm for computing the depth of the tree is provided. Because operations for SOLAP such as drilldown and rollup only need to be sent to a part of the whole network, it is costly for all the nodes in the network to participate in transmitting the operation. In this paper, a multi-granularity index is provided to deal with this problem. The most important computation processing for SOLAP is the aggregation of the sensor data. If it is done by a centralizedway, too much data have to be transmitted by each node in the network and this would cost so the energy constrained sensor nodes in the network so much energy. What is more, this would cause to much data to be lost through the transmission. Innetwork aggregation is obviously a good choice to make this computation efficient. In this paper, a new time scheduling method for innetwork aggregation is provided. Operations for SOLAP should be sent properly to the to certain nodes just as the network is computing. To accomplish this, a time scheduling process for transmitting the operations is provided in this work. For nodes to store multi-granularity partial results of the innetwork aggregation, a data structure is provided. The operations on the structure to accomplish drilldown and rollup are also showed. Using the method above, a prototype for SOLAP is also implemented.