Routing in Multi-Radio, Multi-Hop Wireless Mesh NetworksRichard Draves Jitendra Padhye Brian ZillMicrosoft Research{richdr, padhye, bzill}@microsoft.comABSTRACTWe present a new metric for routing in multi-radio, multi-hop wireless networks. We focus on wireless networks withstationary nodes, such as community wireless networks.The goal of the metric is to choose a high-throughput pathbetween a source and a destination. Our metric assignsweights to individual links based on the Expected Trans-mission Time (ETT) of a packet over the link. The ETTis a function of the loss rate and the bandwidth of the link.The individual link weights are combined into a path metriccalled Weighted Cumulative ETT (WCETT) that explicitlyaccounts for the interference among links that use the samechannel. The WCETT metric is incorporated into a rout-ing protocol that we call Multi-Radio Link-Quality SourceRouting.We studied the performance of our metric by implement-ing it in a wireless testbed consisting of 23 nodes, eachequipped with two 802.11 wireless cards. We find that ina multi-radio environment, our metric significantly outper-forms previously-proposed routing metrics by making judi-cious use of the second radio.Categories and Subject DescriptorsC.2.1 [Network Architecture and Design]: WirelessNetworksGeneral TermsMeasurement, Performance, ExperimentationKeywordsWireless multi-hop networks, multi-radio, routing, perfor-mance1. INTRODUCTIONRouting in ad-hoc wireless networks has been an activearea of research for many years. Much of the original workin the area was motivated by mobile application environ-ments, such as battlefield ad-hoc networks. The primaryPermission to make digital or hard copies of all or part of this work forpersonal or classroom use is granted without fee provided that copies arenot made or distributed for profit or commercial advantage and that copiesbear this notice and the full citation on the first page. To copy otherwise, torepublish, to post on servers or to redistribute to lists, requires prior specificpermission and/or a fee.MobiCom’04, Sept. 26-Oct. 1, 2004, Philadelphia, Pennsylvania, USA.Copyright 2004 ACM 1-58113-868-7/04/0009 ...$5.00.focus in such environments is to provide scalable routing inthe presence of mobile nodes.Recently, interesting commercial applications of multi-hopwireless networks have emerged. One example of such ap-plications is “community wireless networks” [6, 37, 35, 27].Several companies [31, 34] are field-testing wireless networksto provide broadband Internet access to communities thatpreviously did not have such access.In such networks, most of the nodes are either stationaryor minimally mobile and do not rely on batteries. Hence, thefocus of routing algorithms is on improving the network ca-pacity or the performance of individual transfers, instead ofcoping with mobility or minimizing power usage. One of themain problems facing such networks is the reduction in totalcapacity due to interference between multiple simultaneoustransmissions [21].Providing each node with multiple radios offers a promis-ing avenue for improving the capacity of these networks [5,2]. First, it enables nodes to transmit and receive simultane-ously. Otherwise, with only one radio, the capacity of relaynodes is halved. Second, the network can utilize more ofthe radio spectrum. With two radios, a node may transmiton two channels simultaneously. Third, radios that operateon different frequency bands (for example, 802.11a at 5Ghzand 802.11b/g at 2.4Ghz) have different bandwidth, range,and fading characteristics. Using multiple heterogeneous ra-dios offers tradeoffs that can improve robustness, connec-tivity, and performance. Finally, 802.11 radios are off-the-shelf commodity parts with rapidly diminishing prices. Thismakes it natural to consider the use of multiple inexpensiveradios per node.There are other promising approaches for improving thecapacity of multi-hop wireless networks, such as directionalantennas, improved MACs, and channel switching. We be-lieve that these alternative approaches are complementaryto the use of multiple radios. We will discuss this in moredetail in Section 7.When network nodes have multiple radios, the shortest-path algorithm does not perform well. This is illustratedby the following two scenarios. First, consider a network inwhich each node has an 802.11a and an 802.11b radio. Since802.11b radios generally have longer range than 802.11a ra-dios, if we use shortest-path routing, most of the traffic inthe network will be carried over the slower 802.11b links.This is clearly not desirable. Second, consider a network inwhich each node has two 802.11b radios, one tuned to chan-nel 1 and the other tuned to channel 11. Consider a 2-hop(3 nodes) path in this network. A path that is entirely over114channel 1 or 11 will have significantly worse throughput thana path in which the the two hops are on different channels.A shortest-path algorithm that selects a path without ensur-ing that the hops are on different channels will not performwell.What we need is a new routing metric, designed from theground-up for heterogeneous, multi-radio environments. Inthis paper we propose such a metric and present its im-plementation in a protocol that we call Multi-Radio Link-Quality Source Routing (MR-LQSR) protocol. We comparethe performance of our metric against other routing met-rics in a 23-node multi-radio wireless testbed. The resultsfrom the testbed study show that our metric significantlyoutperforms previously-proposed metrics.2. WHY A NEW ROUTING METRIC?Much prior research [43, 4, 24, 13, 18, 20] has recognizedthe shortcomings of shortest-path routing in multi-hop wire-less networks. In this section, we will focus on the ETX(Expected Transmission Count) routing metric proposed byDe Couto et al. [15]. Section 7 discusses other related work.TheworkofDeCoutoet al. shares our goal of using inex-pensive, commodity hardware to build and deploy multi-hopwireless networks. They also use a similar indoor testbed en-vironment with stationary nodes for evaluation. AlthoughETX does very well in homogeneous single-radio environ-ments, as we will show it does not perform as well in envi-ronments with different data rates or multiple radios. Wewill first review the definition of ETX and then discuss itsperformance.The ETX metric measures the expected number of trans-missions, including retransmissions, needed to send a uni-cast packet across