Quality of Service in Mobile Ad Hoc Networks

Abstract

Mobile ad hoc networks (MANETs) form a class of multi-hop wireless networks that can easily be deployed on-the-fly. These are autonomous systems that do not require existing infrastructure; each participating node in the network acts as a host as well as a packet-forwarding router. In addition to the difficulties experienced by conventional wireless networks, such as wireless interference, noise and obstructions from the environment, hidden/exposed terminal problems, and limited physical security, MANETs are also characterized by dynamically changing network topology and energy constraints. While MANETs were originally designed for use in disaster emergencies and defense-related applications, there are a number of potential applications of ad hoc networking that are commercially viable. Some of these applications include multimedia teleconferencing, home networking, embedded computing, electronic classrooms, sensor networks, and even underwater surveillance. The increased interest in MANETs in recent years has led to intensive research efforts which aim to provide quality of service (QoS) support over such infrastructure-less networks with unpredictable behaviour. Generally, the QoS of any particular network can be defined as its ability to deliver a guaranteed level of service to its users and/or applications. These service requirements often include performance metrics such as throughput, delay, jitter (delay variance), bandwidth, reliability, etc., and different applications may have varying service requirements. The performance metrics can be computed in three different ways: (i) concave (e.g., minimum bandwidth along each link); (ii) additive (e.g., total delay along a path); and (iii) multiplicative (e.g., packet delivery ratio along the entire route). While much effort has been invested in providing QoS in the Internet during the last decade, leading to the development of Internet QoS models such as integrated services (IntServ) (Braden, 1994) and differentiated services (DiffServ) (Blake, 1998), the Internet is currently able to provide only best effort (BE) QoS to its applications. In such networks with predictable resource availability, providing QoS beyond best effort is already a challenge. It is therefore even more difficult to achieve a BE-QoS similar to the Internet in networks like MANETs, which experience a vast spectrum of network dynamics (such as node mobility and link instability). In addition, QoS is only plausible in a MANET if it is combinatorially stable, i.e., topological changes occur slow enough to allow the successful propagation of updates throughout the network. As such, it is often debatable as to whether QoS in MANETs is just a myth or can become a reality.