A Comprehensive Analysis of an Adaptive Opportunistic Routing with Congestion Diversity in Wireless Networks

Traditional routing strategies for multi-hop wireless networks forward packets by selecting at the sender side the next hop for each packet. We consider the problem of routing packets across a multi-hop network consisting of multiple sources of traffic and wireless links while ensuring bounded expected delay. Each packet transmission can be overheard by a random subset of receiver nodes among which the next relay is selected opportunistically. The main challenge in the design of minimumdelay routing policies is balancing the trade-off between routing the packets along the shortest paths to the destination and distributing traffic according to the maximum backpressure. Combining important aspects of shortest path and backpressure routing, this paper provides a systematic development of a distributed opportunistic routing policy with congestion diversity (D-ORCD). D-ORCD uses a measure of draining time to opportunistically identify and route packets along the paths with an expected low overall congestion. DORCD is proved to ensure a bounded expected delay for all networks and under any admissible traffic. Furthermore, this paper proposes a practical implementation which empirically optimizes critical algorithm parameters and their effects on delay as well as protocol overhead. Realistic Qualnet simulations for 802.11-based networks demonstrate a significant improvement in the average delay over comparative solutions. I.Introduction Wireless networks typically use routing techniques similar to those in wired networks [15, 16, 9, 4, 5]. These traditional routing protocols choose the best sequence of nodes between the source and destination, and forward each packet through that sequence. Opportunistic routing for multi-hop wireless ad-hoc networks has long been proposed to overcome deficiencies of conventional routing [1]–[5]. Opportunistic routing mitigates the impact of poor wireless links by exploiting the broadcast nature of wireless transmissions and the path diversity. More precisely, the routing decisions are made in an online manner by choosing the next relay based on the actual transmission outcomes as well as a rank ordering of neighboring nodes. The authors in [4] provided a Markov decision theoretic formulation for opportunistic routing and a unified framework for many versions of opportunistic routing [1]–[3], with the variations due to the authors’ choices of costs. In particular, it is shown that for any packet, the optimal routing decision, in the sense of minimum cost or hop-count, is to select the next relay node based on an index. This index is equal to the expected cost or hop-count of relaying the packet along the least costly or the shortest feasible path to the destination. When multiple streams of packets are to traverse the network, however, it might be necessary to route some packets along longer or more costly paths, if these paths eventually lead to links that are less congested. More precisely, and as noted in [6], [7], the opportunistic routing schemes in [1]–[5] can potentially cause severe congestion and Undeti-International Journal of Computer Science information and Engg., Technologies ISSN 2277-4408 || 01062014-015 IJCSIET-ISSUE4-VOLUME2-SERIES2 Page 2 unbounded delays (see examples given in [6]). In contrast, it is known that an opportunistic variant of backpressure [8], diversity backpressure routing (DIVBAR) [7] ensures bounded expected total backlog for all stabilizable arrival rates. To ensure throughput optimality (bounded expected total backlog for all stabilizable arrival rates), backpressure-based algorithms [7], [8] do something very different from [1]–[5]: rather than using any metric of closeness (or cost) to the destination, they choose the receiver with the largest positive differential backlog (routing responsibility is retained by the transmitter if no such receiver exists). This very property of ignoring the cost to the destination, however, becomes the bane of this approach, leading to poor delay performance in low to moderate traffic (see [6]). Other existing provably throughput optimal routing policies [9]–[12] distribute the traffic locally in a manner similar to DIVBAR and, hence, result in large delay. Recognizing the shortcomings of the two approaches, researchers have begun to propose solutions which combine elements of shortest path and backpressure computations [7], [15], [16]. In [7], EDIVBAR is proposed: when choosing the next relay among the set of potential forwarders, EDIVBAR considers the sum of the differential backlog and the expected hop-count to the destination (also known as ETX). However, E-DIVBAR does not necessarily result in a better delay performance than DIVBAR. Instead of a simple addition used in EDIVBAR, this paper provides a distributed opportunistic routing policy with congestion diversity (DORCD) under which the congestion information is integrated with the distributed shortest path computations of [4]. In our previous work [13], ORCD, a centralized version of D-ORCD, is shown to be throughput optimal without discussion on system implications. In this paper, we extend the throughput optimality proof for the distributed version and discuss implementation issues in detail. We also tackle some of the system level issues observed in realistic settings via detailed Qualnet simulations. We then show that DORCD exhibits better delay performance than state of the art routing policies, namely, EXOR, DIVBAR and E-DIVBAR. Before we close, we emphasize that some of the ideas behind the design of D-ORCD have also been used as guiding principles in many routing solutions: some in opportunistic context [14], [15] and some in conventional context [16]. we detail the similarity and differences between these solutions and our work for the sake of completeness, even though, in our study, we have chosen to focus only on solutions with comparable overhread and similar degree of practicality. In [14], perhaps the most related work to ours, the authors consider a flow-level model of the network and propose a routing policy referred to as min-backlogged-path routing, under which the flows are routed along the paths with minimum total backlog. In this light, D-ORCD can be viewed as a packet-based version of the minbacklogged-path routing without a need for the enumeration of paths across the network and costly computations of total backlog along paths. In [15], authors propose a modified version of backpressure which uses the shortest path information to minimize the average number of hops per packet delivery, while keeping the queues stable. In [16], a modified throughput optimal backpressure policy, LIFOBackpressure, is proposed using LIFO discipline at layer 2. Neither of these approaches lend themselves to practical implementations: [15] requires maintaining large number of virtual queues at each node increasing implementation complexity, while [16] uses atypical LIFO scheduler resulting in significant reordering of packets. Furthermore, while LIFOBackpressure policy guarantees stability with minimal queuelength variations, realistic bursty traffic in large multi-hop wireless networks may result in queue-length variations and unnecessarily high delay.