Efficient data management in wireless sensor networks using peer-to-peer techniques

Abstract

The data-centric nature of Wireless Sensor Networks (WSNs) and the severe resource constraints of the sensor nodes distinguish sensor data management from other communication networks. Both the client-server approach as well as the end-to-end communication principle that have been proposed in the for Mobile Ad Hoc Networks (MANETs) do not suit the characteristics of WSNs. Efficient lookup and routing of sensor data are of great significance for WSNs, especially as the size of these networks continues to grow. On an abstract level, structured peer-to-peer protocols, which rely on Distributed Hash Tables (DHTs) provide O(1) complexity for storing and retrieving data in the network, seem overcome these restrictions. However, the fact that they rely on underlayer routing techniques leads to inefficient resource usage in the context of sensor networks. The combination of DHTs and underlayer routing led to the establishment of so called virtual coordinate routing techniques. Most of these algorithms are quite complicated and do not guarantee packet delivery on the shortest path. Additionally, only few of them are implemented in real sensor nodes. In this dissertation, we present the Virtual Cord Protocol (VCP), a virtual position based routing protocol that also provides means for data management such as identifying, storing, and retrieving data items. The key contributions of this protocol are independence of real location information by relying on virtual relative positions of neighboring nodes, the simplicity of obtaining the virtual positions, near optimal routing paths, and high scalability because only information about direct neighbors is needed for routing. Furthermore, VCP provides a unique position for each node and inherently prevents dead-ends. We extensively evaluated VCP in a number of simulation experiments. The simulation results show that VCP consistently provides high throughput and low overhead for a wide range of application scenarios. We compared VCP with Dynamic MANET on Demand (DYMO), a typical MANET protocol, and with Virtual Ring Routing (VRR), another virtual coordinate based approach. In static networks, both VCP and VRR clearly outperform DYMO. In the case of frequent node failures, however, VCP benefits from its light-weight design. Our protocol is more failure tolerant compared to VRR. We finally integrated data replication techniques that support high success rates even in very unreliable sensor networks. A prototype implementation on real sensor nodes outlines the feasibility of our approach in a proof-of-concept study