Elsevier

Computer Networks

Volume 56, Issue 5, 30 March 2012, Pages 1568-1583
Computer Networks

An experimental evaluation of cross-layer routing in a wireless mesh backbone

https://doi.org/10.1016/j.comnet.2012.01.004Get rights and content

Abstract

Over the latest few years, cross-layer design in wireless networks has drawn great attention from the research community. One of the main arguments in favor of such techniques is that the hop-count metric alone is not enough to capture the specificities of wireless links (e.g., interferences, collisions, fading). In this paper, we address a simple yet fundamental question: What are the real improvements that cross-layering can bring to routing performance when compared to the simple hop-count metric? In our experiments, we consider the backbone of a real wireless mesh network composed of 12 routers deployed in an office building. We focus on the stability of routes and their persistence. In spite of the nature of cross-layer metrics that take into account information from different layers, lets them be very reactive to changes, we observe that using these metrics, pairs of nodes tend to mainly use the same set of two or three routes between them.

Introduction

Wireless multi-hop networks consist of a set of nodes (potentially mobile) that communicate through the wireless medium. In such networks, intermediate nodes relay packets in order to reach the final destination. The ordered list of wireless nodes traversed by packets is referred to as a path or route. Selecting the best route between two nodes is the role of the routing protocol. The main question to be answered, in this context, is how to characterize the “quality” of routes so that the best one can be selected. In a multi-hop context, the easiest way to define a distance between two nodes is to use the shortest hop-count metric, which consists in selecting the best route as the route with the minimum number of hops.

The shortest hop-count metric is more suitable in wired networks, where links are more stable, than in wireless networks due to the specificities of these latter (i.e., limited bandwidth, highly dynamic topology, link interference, limited range of links, and broadcast nature of the medium). Couto et al. [1] show through experimental evaluations that routes with minimum hop-count are usually not the best routes in terms of throughput. This is justified by the work of Lundgren et al. [2], who define a specific geographic location, called “communication gray zone”, in which nodes can be sensed by broadcast routing messages. Nevertheless, they cannot relay traffic at rates higher than the broadcast rate. Fig. 1 shows an example of this phenomenon using IEEE 802.11b wireless cards. Given that nodes r01 and r03 sense each other through broadcasts sent at 1 Mbps, data exchanged between them will never use a transmission rate higher than 1 Mbps. However, using the two-hop route through node r02, the end-to-end transmission rate can be improved to 5.5 Mbps (two transmissions at 11 Mbps). In addition to the gray zone phenomenon, Heusse et al. [3] discuss the performance anomaly of IEEE 802.11b. They observe that the throughput of nodes transmitting at high rate decreases when one or more nodes in their range transmit at a lower rate. For a more complete view of cross-layering techniques for wireless multi-hop networks, the reader can refer to work of Borges et al. [4] that present a taxonomy of routing metrics based on cross-layer techniques.

For the aforementioned reasons, extensive research on routing solutions based on cross-layer techniques has been conducted in the latest years. Awerbuch et al. [5] propose a new metric named Medium Time Metric (MTM) that gives a weight to each link relative to the packet transmission time on the medium. Karbaschi et al. [6] propose a similar approach, where the number of retransmission done at MAC layer is used instead of packet transmission time. Chiang et al. [7] tackle the congestion problem of wireless links by proposing a cross-layer approach where the transport layer interacts with the physical layer. Iannone et al. [8] propose a cross-layer metric that takes into account the transmission power level, transmission rate, and interference. As a matter of fact, many proposals in literature rely on complex theoretical formalization and/or simulation analysis. They can be hardly implemented in today’s system or cannot be supported by all wireless cards, as shown by Ben Abdesslem et al. [9].

Despite the proposal of a large number of routing protocols (including several cross-layer based approaches), relatively little research work has been done on how to take into account the notion of link stability in route selection algorithms. Some routing protocols (mainly reactive protocols where route calculation is expensive), such as SSA (Signal Stability based Adaptive routing) and ABR (Associativity-Based Routing) use models based on signal strength or pilot signal to estimate the link stability [10], [11], [12]. Paul et al. [13] proposed the “affinity” parameter that consists in predicting link lifetime. This affinity parameter is based on the strength and the stability of relationships between a node and its neighbors. Inspired by this idea, Agarwal et al. [14] proposed a routing algorithm that incorporates the link lifetime prediction on the basis of affinity appraisal. In that work, the authors also considered path length while choosing an optimal route for a TCP source. These works are, however, mainly used for on-demand routing protocols, where nodes need to maintain routes during communication as long as possible to avoid delay and overhead during new route discovering. Kakitani et al. [15] proposed a routing metric based on cross-layer approach, where the probability of route failure is calculated by taking into account the physical layer information as the frame error rate (FER). Through simulations, they show that routes are more stable when using a cross-layer metric. Ramachandran et al. [18] study the stability behavior of wireless paths. They collected, through wireless beacons, information about link quality and then computed, on a per-minute basis and offline, all possible routes between sources and destinations using an implementation of the Dijkstra shortest-path algorithm.

In this paper, we investigate through experimentations on a real testbed, and with real routing protocols, the benefits that cross-layer, compared to the hop-count metric, can bring (or not) to the particular case of proactive multi-hop routing protocols in terms of route stability, and show how a cross-layer metric can be sensitive to variations of the environment.1 More specifically, we address the following questions:

  • What is the proportion of time during which the dominant route between each pair of nodes is active? Is this dominant route persistent?

  • Are there any sub-dominant routes? If yes, what about their dominance?

  • How long does each one of the routes last? How does the distance between nodes affect their dominance?

In order to answer these questions, we propose a measurement-based study of the benefits provided by cross-layer routing in terms of stability in a real multi-hop wireless network testbed. To this end, we use two well-known and standardized proactive routing protocols, namely DSDV (Destination Sequenced Distance Vector [16]) and OLSR (Optimized Link State Routing protocol [17]). The longest period of time a route is used between two nodes, the more it is considered as stable. To this end, our measurements focus on the dominance and the persistence of dominant and sub-dominant routes, including the evaluation of the behavior of routes when the geographical distance between nodes increases. By injecting artificial traffic, we also evaluate the impact of data traffic on route stability. Our findings show that pairs of nodes tend to use the same routes all the time, even when cross-layer metrics are deployed.

The remainder of the paper is organized as follows. In Section 2, we overview possible cross-layer metrics in wireless routing protocols. In Section 3, we describe our testbed, introduce the measurement setup, and define the scenarios of our experiments. In Section 4, we investigate the dominance of the first most used route between each source–destination pair as well as the persistence of the rest of the most used routes, with and without cross-layer metric. We also evaluate the impact of the physical distance between nodes on route dominance. Before concluding this paper in Section 6, we study in Section 5 the impact of data traffic on cross-layer metrics sensitivity.

Section snippets

Cross-layer routing metrics

Cross-layer design consists in creating new ways of interaction between different layers of the protocol stack. As an example of such an interaction, Fig. 2 shows how different entities at different layers of the protocol stack may interpret differently a single event such as a simple packet loss. This is due to the lack of interaction among different layers. Making possible their interactions in order to allow overcoming such a limitation is what the literature refers to as cross-layering.

The

Experimentation setup

Our evaluation is based on routing information collected from a real testbed (namely MeshDVnet). We start this section by briefly describing this testbed and then discuss the topology of the network as well as the experimentation scenarios. We explain how we extract route information after collecting all routing tables in a central database. We finally introduce the routing protocol daemons used during the experimentation and the interactions among layers in route selection.

Route stability

In this section, we evaluate how cross-layer design in proactive wireless routing protocols is sensitive to wireless environment changes in terms of route stability. As highlighted before, the longest consecutive period of time a route is used between two nodes, the more it is considered as stable, i.e., route oscillation is equivalent to instability. In order to isolate the stability of the single metric with regard to the wireless environment, we first measure routing changes without

Traffic performance

As mentioned before, nodes that are too far away may only be detected when using the lowest transmission rate (larger range). In this case, they cannot relay traffic at rates higher than the broadcast rate. To go deeper into our study, we re-perform the same measurement campaign described in Section 3 while generating different types of traffic between nodes during three consecutive hours.

For DSDV protocol, we start by generating ICMPv6 traffic using the Ping6 tool between node r09 and r08

Conclusion

In this paper, we investigated the benefits of cross-layer design in wireless multi-hop routing protocol, and showed how cross-layer metrics can be sensitive to the medium’s conditions changes in terms of stability of routes in a real wireless multi-hop network. For our study, we used DSDV and OLSR as routing protocols with hop-count and cross-layer metrics (Tx rate and ETX). We observed that pairs of nodes tend to use the same routes between them, even using cross-layer metrics, despite nature

Mehdi Bezahaf is a postdoctoral fellow at the LIP6 laboratory of UPMC Sorbonne Universités (France). He received his M.Sc. and Ph.D. Degrees from UPMC Sorbonne Universités in 2007 and 2010 respectively. His research interests include mobility management architecture, IPv6 auto-configuration, distributed hash tables, and multi-hop wireless networks. For more information, visit http://www-npa.lip6.fr/bezahaf.

References (41)

  • V.C. Borges et al.

    Cross-layer routing metrics for mesh networks: current status and research directions

    Computer Communications

    (2011)
  • D.S.J.D. Couto et al.

    Performance of multihop wireless networks: shortest path is not enough

    ACM SIGCOMM Computer Communication Review

    (2003)
  • H. Lundgren, E. Nordströ, C. Tschudin, Coping with communication gray zones in ieee 802.11b based ad hoc networks, in:...
  • M. Heusse, F. Rousseau, G. Berger-Sabbatel, A. Duda, Performance anomaly of 802.11b, in: IEEE Infocom, San...
  • B. Awerbuch et al.

    The medium time metric: high throughput route selection in multi-rate ad hoc wireless networks

    Mobile Networks and Applications

    (2006)
  • G. Karbaschi, A. Fladenmuller, A link quality and congestion-aware cross layer metric for multi-hop wireless routing,...
  • M. Chiang

    To layer or not to layer: balancing transport and physical layers in wireless multihop networks

    IEEE Journal on Selected Areas in Communications

    (2005)
  • L. Iannone, R. Khalili, K. Salamatian, S. Fdida, Cross-layer routing in wireless mesh networks, in: International...
  • F. Ben Abdesslem, L. Iannone, M.D. de Amorim, K. Kabassanov, S. Fdida, On the feasibility of power control in current...
  • R. Dube et al.

    Signal stability based adaptive routing (ssa) for ad-hoc mobile networks

    IEEE Personal Communications

    (1997)
  • C.-K. Toh

    Associativity-based routing for ad-hoc mobile networks

    Wireless Personal Communications

    (1997)
  • G. Lim, K. Shin, S. Lee, H. Yoon, J.S. Ma, Link stability and route lifetime in ad-hoc wireless networks, in:...
  • K. Paul, S. Bandyopadhyay, A. Mukherjee, D. Saha, Communication aware mobile hosts in ad-hoc wireless network, in: IEEE...
  • S. Agarwal, A. Ahuja, J.P. Singh, R. Shorey, Route-lifetime assessment based routing (rabr) protocol for mobile ad-hoc...
  • M. Kakitani, A. Munaretto, M. Fonseca, An optimal path stability achieved by a cross-layer proposal for wireless...
  • C. Perkins, P. Bhagwat, Highly dynamic destination-sequenced distance-vector routing (dsdv) for mobile computers, in:...
  • P. Jacquet, P. Mühlethaler, T. Clausen, A. Laouiti, A. Qayyum, L. Viennot, Optimized link state routing protocol for ad...
  • K. Ramachandran, I. Sheriff, E. Belding, K. Almeroth, Routing stability in static wireless mesh networks, in: Passive...
  • D. Mitra et al.

    Virtual Private Networks: Joint Resource Allocation and Routing Design

    (1999)
  • C.R. Michael Neely, Eytan Modiano, Dynamic power allocation and routing for time varying wireless networks, in: IEEE...
  • Cited by (0)

    Mehdi Bezahaf is a postdoctoral fellow at the LIP6 laboratory of UPMC Sorbonne Universités (France). He received his M.Sc. and Ph.D. Degrees from UPMC Sorbonne Universités in 2007 and 2010 respectively. His research interests include mobility management architecture, IPv6 auto-configuration, distributed hash tables, and multi-hop wireless networks. For more information, visit http://www-npa.lip6.fr/bezahaf.

    Luigi Iannone is currently Senior Research Scientist at TU Berlin/Deutsche Telekom Laboratories AG (T-Labs) since September 2008. He worked as post-doc at Université catholique de Louvain (UCL - Belgium) in 2007/2008. He holds a Ph.D. in Computer Science from UPMC Sorbonne Universités, France (2002/2007) and a degree in Computer Engineer from Universitá degli Studi di Pisa - Italy (2001). His current research interests include intra- and interdomain routing, mobility, and wireless networks.

    Marcelo Dias de Amorim is a CNRS permanent researcher at the computer science laboratory (LIP6) of UPMC Sorbonne Universités, France. His research interests focus on the design and evaluation of dynamic networks as well as service-oriented architectures. For more information, visit http://www-npa.lip6.fr/amorim.

    Serge Fdida is a professor at UPMC Sorbonne Universités, France, since 1995. He received his Ph.D. in 1984, and the Habilitation Diriger des Recherches (HDR) in Modelling of Computer Networks in 1989 from the same University. His research interests are in the area of high speed and mobile networking, pervasive communication, resource management and performance analysis. He is heading the Network and Performance Analysis group of the computer science laboratory (LIP6).

    View full text