Data Dissemination in Wireless Networks with Network Coding

Abstract

Abstract-We investigate the use of network coding for information dissemination over a wireless network. Using network coding allows for a simple, distributed and robust algorithm where nodes do not need any information from their neighbors.In this paper, we analyze the time needed to diffuse information throughout a network when network coding is implemented at all nodes. We then provide an upper bound for the dissemination time for ad-hoc networks with general topology. Moreover, we derive a relation between dissemination time and the size of the wireless network. It is shown that for a wireless network with N nodes, the dissemination latency is between O(N ) and O(N 2 ), depending on the reception probabilities of the nodes. These observations are validated by the simulation results. I. INTRODUCTIONThe information dissemination problem, at its root, is a classical broadcast problem: sharing data residing at one node (source) with all others (destinations) in the network. Of late, a more modern version of the one-to-many (broadcast) problem has gained prominence; this is typically referred to as data sharing among multiple peer-to-peer (p2p) nodes, or the allto-all problem. This problem arises when each node in a network obtains only a fraction of the total information (e.g., part of a video-on-demand file or a software update) desired collectively by all. In a simplified version of the all-to-all data-dissemination problem, a source file desired by all is divided into N mutually exclusive information packets, and each packet is stored at exactly one node in the network [1], [2]. Every node's objective is to acquire the remaining N − 1 pieces of the source file; the order in which each node receives the remaining information packets is not relevant.Traditionally, the data dissemination problem over decentralized network architecture has focused on the impact of the dissemination algorithm designed to optimize a performance metric, such as dissemination latency (i.e., the time required for all nodes to acquire the entire file [3], [4]). Authors in [5], [6] showed that using Network Coding (NC) for dissemination in a wired network can improve dissemination latency.Recently, authors in [7] used NC to diffuse information in an ad-hoc wireless network. However, their analytical models have largely suffered from unrealistic assumptions that are unsuited to a wireless network-notably that of pure fail/success (0 − 1)-whereby each transmission is either successfully received by all neighbors or fails. Our analysis advances the state of the art by using a more appropriate link model whereby, for each broadcast, sink nodes successfully receive the transmitted packet with a reception probability that is dependent upon the nodes' respective locations. We provide an upper bound

Performance of Random Linear Coding Over Multiple Error-Prone Wireless Links

“…In this paper we have assumed that all sources are synchronised and therefore send the same block; we will study the influence of considering scenarios where each of the transmitters could send different blocks. This will be as well extended so as to consider multiple destinations, using scenarios similar to those studied by Firooz et al [6].…”

Implementation of Network-Coded Cooperation for Energy Efficient Content Distribution in 5G Mobile Small Cells

“…The eNodeB (eNB) unicasted parts of the anchors to the UE, which broadcasted the encoded anchors. Firooz et al [42] studied the latency upper and lower bounds for content distribution using network coding, and Torre et al [43] implemented and studied the latency of the scenario. The analytical model of RLNC has been studied in [44], who proposed a Newton method approximation to adjust the channel losses to optimize channel usage.…”

Sparse Random Linear Network Coding For Low Latency Allcast

“…In [7], an RLNC allcast system is analysed, on graphs where edges denote erasure channels, and the medium is broadcast in nature. A multiple access system is in operation, restricting nodes to broadcast one at a time, and ensuring that nodes gain channel access with equal probability (modelling CSMA/CA).…”