Cooperative Data Exchange in Broadcast Networks

Abstract Network coding is a new discipline that encompasses a broad range of information processing techniques beyond traditional channel coding. Initially, network coding was proposed as a way to maximize the network throughput by combining packets that belong to different flows. Over the last two decades, many novel network coding techniques have been proposed for wireless and wireline networks as well as for distributed storage settings. It was shown that network coding can significantly improve the performance, robustness, and security of communication networks and storage systems. The chapter discusses the applications of network coding for efficient data exchange in one-hop wireless networks, focusing on the cooperative data exchange (CDE) problem and its variations. An instance of the CDE problem includes a set of wireless clients and a set of information messages such that each client initially holds a subset of the messages in the set. The clients use a shared lossless broadcast channel to exchange coded combinations of the messages they hold. The goal is to enable every client to recover all missing messages while minimizing the overall number of transmissions. This problem arises naturally in many practical settings, including synchronization of the local databases and file systems, gathering data from correlated sources in sensor networks, and replicating data to increase reliability. We begin by presenting a randomized algorithm for Problem CDE and analyzing its correctness and performance. Next, we discuss coding schemes that can tolerate a limited number of client failures. Resilience to failures is achieved by proactively adding redundant transmissions that ensure that all valid clients are able to receive all messages in any failure scenario. Because the problem of designing a failure-resilient coding scheme that minimizes the total number of transmissions is intractable, we present an approximation algorithm with provable performance guarantees. Finally, we consider settings in which the data exchange is performed in the presence of a wiretapper whose goal is to obtain information about the original message set.