|
Inter-Networking Research Group (i-NRG) |
|
|
Inter-Networking Research Group (i-NRG) |
A multi-hop, wireless ad hoc network (or MANET) operates without any fixed infrastructure. Each host communicates with each other through wireless packet radios. Because of the limited radio propagation range, routes can often be multi-hop. Hence, every host may act as a packet forwarder as well as source or destination of traffic. The types of scenarios targeted by MANETs make group-oriented services such as data dissemination and teleconferencing a key application domain. In particular, the mission-critical characteristics of a number of these applications MANETs (e.g., emergency response, special civilian or military operations) call for efficient reliable multi-point communication protocols. Undoubtedly, ``network-supported'' multicast communication is an efficient means of supporting group-oriented applications. This is especially true in MANETs where nodes are energy- and bandwidth limited.
Several features unique to MANETs make the design of MANET reliable multicast transport mechanisms quite challenging. Among these features, we highlight: (1) MANET's heterogeneous loss characteristics due to factors such as mobility, node density, time-varying channel conditions, (2) effects of lower layer protocols, e.g., inherent unfairness and unreliability of contention-based medium access control protocols (e.g., IEEE802.11 uses plain CSMA when broadcasting packets and thus do not provide reliable broadcast delivery), and (3) MANET's extreme sensitivity to offered load. These MANET features render design choices used in reliable multicast protocols for wired networks not at all applicable to MANET environments.
We introduce Reliable Adaptive Congestion-controlled transport Protocol (ReACT)~\cite{react, react-tr}, a novel reliable multicast transport protocol for MANETs. Some of the distinguishing features of ReACT are that (1) it combines source-based rate control with local error recovery and (2) uses loss differentiation to trigger either source-based control or local recovery. The goal is to recover from localized losses (e.g. due to node mobility, link quality, channel contention) using nearby group members, while congestion losses are reported to the source, triggering error- as well as congestion recovery. Through extensive simulations, we evaluate ReACT's performance under a variety of MANET scenarios, including different offered load and mobility conditions, and compare it against a strictly end-to-end (i.e., no localized recovery) scheme. Our results show that ReACT is the best performer in terms of reliability. In our future we intend to evaluate the impact of providing reliability at the link level.
