R-Code: Network coding-based reliable broadcast in wireless mesh networks

Abstract

Broadcast is an important communication primitive in wireless mesh networks (WMNs). Applications like network-wide software updates require reliable broadcast to ensure that every node in the network receives the information completely and correctly. With underlying unreliable wireless links, a key challenge in implementing reliable broadcast in WMNs is to achieve 100% information reception rate at every node with high communication efficiency and low latency. Recently, network coding has emerged as a promising coding scheme in terms of communication efficiency especially for one to many communication patterns. In this paper, we put forward R-Code, a network coding-based reliable broadcast protocol. We introduce a guardian–ward relationship between neighboring nodes that effectively distributes the responsibility of reliable information delivery – from the global responsibility of the source to the localized responsibilities of guardians to their corresponding wards. We use a link quality-based minimum spanning tree as a backbone to guide the selection of guardians adaptively and the transmission of coded packets accordingly. Opportunistic overhearing is also utilized to improve the performance of the protocol. Extensive simulation results show that R-Code achieves 100% packet delivery ratio (PDR), while enjoying significantly less transmission overhead and shorter broadcast latency, compared with a state-of-the-art reliable broadcast protocol, AdapCode.

Introduction

Wireless mesh network (WMN) emerges as a promising technique to provide high-bandwidth Internet access to a large number of mobile devices in a specific area. Broadcast is an important function in WMNs. For example, it is necessary for software updates which may happen at the initial deployment and maintenance period, or is used in multimedia services like video/audio downloading. A key requirement of these applications is to strictly guarantee 100% packet delivery ratio (PDR), which means every node has to download every bit of the broadcasted file. In addition, efficiency is another important concern. Since other normal unicast traffics may exist in the network at any time, broadcast applications should have good coexistence with these traffics, which translates into consuming minimal amount of network bandwidth and disseminating the file with low latency.

It is nontrivial to design an efficient reliable broadcast protocols for real WMNs. The fundamental challenge comes from the unreliable nature of the wireless link [2], which is due to packet losses caused by channel fading and interferences. In order to guarantee 100% PDR with those unreliable links, some previous schemes [3], [4] use automatic repeat request (ARQ) technique, which requires the receivers to provide explicit feedbacks of the packet reception status to the source. However, this will cause “ACK implosion” problem which may incur a large amount of redundant transmissions. Other schemes [5], [6], [7] combine ARQ with forward error correction (FEC) technique to reduce the transmission overhead while still guaranteeing 100% PDR. Yet, these technique consider the wireless link as point-to-point, and neglect the fact that wireless medium is broadcast in nature. This leads to duplicate transmissions at intermediate nodes, which are not efficient enough.

Recently, network coding (NC) has been proposed as an effective technique to increase the network bandwidth efficiency [8]. In contrast to FEC, NC gives intermediate nodes the ability of randomly encoding different packets received previously into one output packet. Thus, although multiple intermediate nodes may receive the same packet, they will broadcast different re-coded packets that are linearly independent with each other with high probability. Each of these re-coded packets can benefit other nodes that overhear it, which avoids the duplicate transmissions. Theoretical analysis has demonstrated that NC is able to approach the multicast and broadcast capacity in multi-hop wireless networks [9], [10]. Due to the high complexity of implementing network coding, practical NC-based broadcast schemes have also been proposed [11], [12], [13], [14], [15], [16], where NC is shown to have a noticeable gain in bandwidth efficiency. However, most of these schemes only provide reliability with best effort rather than guaranteeing 100% PDR, with the exception of AdapCode [13] and Pacifier [16]. However, AdapCode is purposefully designed for wireless sensor networks and is not efficient when directly applied into WMNs. While MORE and Pacifier focuses on multicast in WMNs, they do not consider to exploit the specific characteristic of broadcast, that is, every node has to receive the whole file. In fact, when a node has received a certain amount of information, it can be regarded as a temporary source, which can guarantee the reliable reception of its neighbors in an efficient way.

In this paper, we propose R-Code, an efficient distributed Reliable broadcast protocol in WMNs based on network Coding which guarantees 100% PDR. The core idea of R-Code is to establish a guardian–ward relationship between neighboring nodes, so that the global responsibility of the source to ensure the reliable reception of all the nodes in the network is distributed to all the guardians. This is because a guardian is a temporary source that is much closer to its wards than the original source, thereby it can guarantee their reliable reception of the file more efficiently. A link quality-based minimum spanning tree is constructed to guide the selection of guardians and packet transmissions accordingly. A guardian is the best node in a ward’s neighborhood to ensure the reliable reception of the ward with the least number of transmissions. The guardian–ward relationship is adaptively maintained throughout the broadcast session to exploit the benefit of opportunistic overhearing. In addition, intra-flow NC is adopted to further reduce the total number of transmissions and simplify the coordination between multiple transmitters. Moreover, R-Code applies a source rate limiting mechanism to alleviate the collisions in the network. We evaluate R-Code and compare it with AdapCode by extensive simulations. The simulation results show that R-Code uses up to 15% less number of transmissions and 65% shorter broadcast latency than that in AdapCode to disseminate the same file.

The rest of the paper is organized as follows: we give related work in Section 2. In Section 3, we describe the preliminaries. The analysis of existing schemes is presented in Section 4. In Section 5, we introduce the design of R-Code protocol in detail. Section 6 presents the simulation results and Section 7 wraps up the paper.