|Abstract:||A mobile ad hoc network (MANET) is formed by a group of mobile nodes connected by wireless links. The nodes can talk to each other by direct peer-to-peer wireless communication when they are close to each other. When the sender and receiver are far away, their packets can be forwarded by the intermediate nodes along a multi-hop path. As an emerging networking technique, a MANET is envisioned to become a stand-alone network for a group of mobile users, or as a stub-network to connect to the Internet.
In this thesis, we focus on the flow control problem in this network. Depending on whether the routers agree to forward packets for each other, different flow control schemes are proposed for both cooperative and non-cooperative MANET environments. In cooperative flow control, we first study the two prominent flow control schemes in the Internet, i.e., TCP and equation-based flow control. We provide an augmentation to TCP in order to improve its performance, and discover several limitations of applying equation-based flow control to the MANET domain. Prompted by the deficiencies of these implicit flow control schemes, we design a router-assisted explicit rate-based flow control scheme (called EXACT) for MANETs. Our result shows that EXACT achieves much better performance than the implicit flow control schemes. Therefore, we advocate using router-assisted explicit flow control, as exemplified by EXACT, in the MANET domain.
Traditionally, flow control has been studied in a cooperative network environment. Such cooperative behavior is not a realistic assumption in a public MANET formed by a random group of strangers. Users in this network are likely to behave selfishly, by refusing to forward other users' packets. Under this non-cooperative environment, flow control and incentive engineering are two problems closely related to each other. A flow control solution would be meaningless if the intermediate routers do not agree to carry the traffic for others. However, existing work in incentive engineering provides only stand-alone solutions, and fails to recognize its close relation with flow control. To this end, we propose a scheme called iPass, which is a joint solution for flow control and incentive engineering in MANETs. iPass adopts the ``pay for service'' model of cooperation, and utilizes an auction mechanism at each router to fairly allocate bandwidth resources to the passing flows. At the same time, the auction market gives explicit rate signals to the end hosts for flow control purposes. By design of the auction algorithm, iPass achieves several desirable system properties, such as bidding user's true valuation for bandwidth. Therefore, iPass provides an innovative and graceful solution for the flow control and incentive engineering problems in a non-cooperative MANET environment.