An Adaptive, High Performance MAC for Long-Distance Multihop Wireless Networks Sergiu Nedevschi UC Berkeley EECS Department Berkeley, CA, USA sergiu@cs.berkeley.edu Rabin K. Patra UC Berkeley EECS Department Berkeley, CA, USA rkpatra@cs.berkeley.edu Sonesh Surana UC Berkeley EECS Department Berkeley, CA, USA sonesh@cs.berkeley.edu Sylvia Ratnasamy Intel Research Berkeley Berkeley, CA, USA sylvia.p.ratnasamy@intel.com Lakshminarayanan Subramanian NYU, CS Department New York,NY, USA lakshmi@cs.nyu.edu Eric Brewer UC Berkeley EECS Department Berkeley, CA, USA brewer@cs.berkeley.edu ABSTRACT We consider the problem of efficient MAC design for long-distance WiFi-based mesh networks. In such networks it is common to see long propagation delays, the use of directional antennas, and the presence of inter-link interference. Prior work has shown that these characteristics make traditional CSMA-based MACs a poor choice for long-distance mesh networks, prompting several recent research efforts exploring the use of TDMA-based approaches to media ac- cess. In this paper we first identify, and then address, several short- comings of current TDMA-based proposals, which exhibit inef- ficienct throughput and delay charactersistics as they use fixed- length transmission slots that cannot adapt to dynamic variations in traffic load. We show that throughput achieved by existing solu- tions falls far short of the optimal achievable network throughput. Current TDMA-based solutions also only apply to bipartitie net- work topologies due to interference scheduling contraints. In this paper, we present JazzyMac, a simple, practical and effi- cient MAC protocol that addresses the above limitations. JazzyMac achieves efficiency by enabling variable-length link transmissions slots; each node can adapt the length of their transmission slots in accordance with changing traffic demands. JazzyMac is practical as it can be applied to arbitrary network topologies, and each node can use purely local information for slot adaptation. Finally, the use of dynamic slot lengths allows JazzyMac to achieve better tradeoffs between throughput and delay. We evaluate JazzyMac using detailed simulation over a range of traffic patterns and realistic topologies. Our results show that JazzyMac improves throughput in all considered scenarios. This improvement is often substantial (e.g.,in 50% of our scenarios, throughput improves by over 40%) and is particularly pronounced for the common case of asymmetric traffic (e.g.,leading to almost 100% improvements). Furthermore, JazzyMac can achieve much better average delay for the same throughput. Permission to make digital or hard copies of all or part of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for profit or commercial advantage and that copies bear this notice and the full citation on the first page. To copy otherwise, to republish, to post on servers or to redistribute to lists, requires prior specific permission and/or a fee. MobiCom’08, September 14–19, 2008, San Francisco, California, USA. Copyright 2008 ACM 978-1-60558-096-8/08/09 ...$5.00. Categories and Subject Descriptors C.2.1 [Computer Systems Organization]: Computer- Communication Networks—Network Architecture and Design General Terms Algorithms, Design, Performance Keywords Multi-hop wireless, Long-distance point-to-point wireless, MAC protocols 1. INTRODUCTION Multi-hop WiFi long-distance networks (WiLD) networks have become increasingly popular in the last few years, providing cost- efficient connectivity to sparsely populated areas and rural regions in developing and industrialized countries alike. Example deploy- ments include the Digital Gangetic Plains project [19], the AirJaldi and Aravind networks [23] and the Akshaya network [24]. These networks serve thousands of users, providing videoconferencing and VoIP services in addition to basic Internet access. Due to the presence of long-distance links and their use of direc- tional antennas, WiLD networks present unique challenges relative to traditional short-range mesh networks [3]. Specifically, these net- works suffer from long propagation delays and an increased like- lihood of inter-packet collisions. In addition, deployments where only a small number of non-overlapping wireless channels are available also suffer from inter-link interference. Prior work has shown that these challenges make traditional MACs based on car- rier sensing, a poor fit for WiLD networks [21]. To address these challenges, several TDMA-based MAC solutions such as 2P [19] and WiLDNet [16] have been developed and are currently used in practical deployments. This paper identifies and addresses certain key performance limitations in 2P and WiLDNet. These limitations arise primarily because these solutions rely on a TDMA schedule with fixed-length slots and hence cannot adapt to dynamic traffic variations. In this paper, we first gauge the potential for improved perfor- mance that might result from a MAC solution that takes advan- tage of observed traffic conditions. For this, we compute the opti- mal throughput achievable in WiLD networks assuming complete knowledge of the network traffic workload (our computation here