Location-centric storage for sensor networks

This paper presents location-centric storage (LCS), a novel distributed data storage protocol for sensor networks. In the protocol, each event detected by sensors is associated with an intensity value (delta)(by sensors), where delta is a parameter that depends on the characteristics of the event and the application context. When event information is broadcast, a sensor decides whether to store the record of an event by checking its distance to the event location and the delta. In general, the higher the intensity of an event, the further its information can propagate geographically in the sensor network. Besides, the closer to the event location, the denser the sensors are that store the event information, and thus the quicker and better a user can know about the event (by reading from surrounding sensors). The protocol utilizes network resource efficiently. In particular, the storage load of sensors is independent of the network size, and is evenly distributed across the network. Moreover, the communication distance for getting event information is small. Therefore, the protocol has great scalability. We provide detailed theoretical analysis and simulation study to support the claims. We also ran simulations to show the advantage of our protocol over some previous work. LCS can be used for applications such as context-dependent information mining in pervasive computing and on-demand warning in surveillance sensor networks