Distributed data storage in sensor networks based on Raptor codes

In this paper an algorithm for distributed data storage in large-scale sensor networks has been proposed. The main objective is to distribute the generated information throughout the network to increase its lifetime. At the same time it is desired that the original data can be recovered later by collecting the contents of a limited number of sensor nodes. In the scenario considered here the sensor nodes have limited energy and memory and each sensor saves only one encoded packet. Also sensors do not hold any routing table and no global information regarding network topology is available. The algorithm proposed here is based on Raptor codes and it inherits their linear time encoding and decoding complexity. Random walks are used for disseminating data amongst sensor nodes. Major benefits of the algorithm presented here is that it utilizes the online decoding property of Raptor codes by achieving the desired code degree distribution and for estimating the required global information regarding network topology, no excessive random walks and transmissions are used.