we analyze TCP performance in asymmetric networks, where the throughput significantly depends on the reverse direction and packet loss. We introduce a new operational model called the "AMP model" which, with an understanding of TCP dynamics and buffer management techniques explains the performance effects seen. We apply this model to guide design improvements in buffer management (ack-regulation) and scheduling schemes to achieve performance improvements of an order of magnitude or more. The schemes have been tested through simulation and implementation, and the module driven improvements observed are better than those proposed in earlier literature.