Distributed data storage in sensor networks using decentralized erasure codes

We assume a large-scale wireless sensor network of n nodes, measuring a physical quantity of interest (e.g. temperature). Most of the sensors are used as storage devices and relays while only a fraction (k out of n) is gathering data. We assume that each node can store at most one data packet. We are interested in storing these k data packets in the n nodes using redundancy so that we can reconstruct the original measurements by asking any k storage nodes in the network (with high probability). We show how this problem is closely related with the distributed construction of an erasure code. To solve this problem, we introduce a new class of erasure codes that can be created without gathering all data at one central location and show that by using these decentralized erasure codes, it is possible to diffuse the data by "pre-routing" only O(ln n) packets per data node to randomly selected storage nodes.