| The prevalence of high bandwidth-delay product networks creates significant challenges for TCP and AQM schemes. To that end, various mechanisms have been proposed to either adaptively adjust sending window size by amending the parameters of Additive-Increase Multiplicative-Decrease, use different congestion signals, or even explicitly signal the congestion information to the sender. Often times, such mechanisms fail to simultaneously achieve high utilization and fairness while maintaining low persistent queue length and minimizing congestion-induced packet drop rate.;In contrast, a pair of recently proposed protocols, XCP and VCP, can achieve near zero congestion loss in wired networks by decoupling fairness control from efficiency. In addition, the operation of any congestion control protocol over wireless networks is subject to a significant performance degradation without differentiating between congestion-caused loss associated with network buffering and error-caused loss associated with fading effects.;This dissertation provides a cross-layer framework of analysis, simulation, implementation, emulation, and performance profiling of congestion control protocols over wireless networks. Relying on simulation, emulation, and implementation, we demonstrate that VCP represents a high performing yet practical congestion control protocol for wireless networks. Further, some of the shortcomings of VCP are identified including its oscillatory behavior in the presence of link estimation errors, relatively low speed of convergence and poor fairness characteristic in moderate bandwidth high delay networks due to utilizing an insufficient amount of congestion feedback.;Next, a distributed congestion control scheme is proposed that distributes and extracts congestion related information into and/or from a chain of packets. Utilizing such a scheme, we design and implement a distributed ECN-based congestion control protocol to which we refer as Multi Packet Congestion Control Protocol (MPCP). In contrast to VCP, MPCP is able to relay a more precise congestion feedback while preserving the use of no more than the two ECN bits in each packet.;Additionally, a heuristic scheme for identifying the cause of loss in wireless network is proposed. This scheme can be applied to any feedback-based congestion control algorithm and can achieve a significant performance improvement over lossy wireless networks. |
