|Ph.D Student||Levin Anna|
|Subject||Mobility Management Schemes for Improving TCP Performance|
in the Next Generation Wireless Networks
|Department||Department of Computer Science||Supervisor||PROF. Reuven Cohen|
|Full Thesis text|
Data applications such as HTTP, FTP and adaptive media streaming require reliable data delivery over the network. TCP is the most widely used transport protocol for this purpose, because its congestion control is essential for guaranteeing network stability, and its flow control and end-to-end reliability remove the need for loss recovery in the Application layer. These issues are especially important in the future wireless networks, which will use wireless technology not only in the last hop, but also in the backbone.
However, TCP introduces significant latency and throughput variability in the presence of mobility and frequent handovers. It was optimized to allow good bandwidth utilization with congestion control in fixed networks. Therefore, in the case of handovers, TCP considers packet losses as a congestion sign, resulting in unnecessary timeouts and retransmissions.
In this thesis we present mobility management schemes aiming at TCP performance enhancement in various types of mobile networks. First, we present two handover schemes: drop and forward. In drop, packets received by the base station after the mobile host has disconnected are dropped, whereas in forward these packets are forwarded to the new base station. We conduct a mathematical analysis of various TCP flavors and compare our findings to simulation results applied to future cellular networks.
One of the main conclusions from the comparison between drop and forward is that the effect of handovers can be mitigated if the packets received by the old base station during handover are forwarded to the new base station. However, if bandwidth on the backbone is scarce, as in the multihop wireless networks, then the bandwidth for packet forwarding is limited. We aim at optimization of bandwidth usage by the forward scheme for adaptive streaming in mobile networks; we define a related optimization problem, and propose efficient algorithms for solving it.
Finally, we propose a new cross-layer mechanism, referred to as Route-control. This mechanism works in the Network and Transport layers and aims at enhancing the performance of reliable Transport protocols in mobile networks. The main idea behind the proposed mechanism is notifying the sender when the packets of a Transport layer flow change their route. We apply the Route-control mechanism to mobile ad-hoc networks and show that the sender can benefit from route change notification when deciding whether to retransmit a missing segment or to wait, when estimating the round trip time, and when deciding whether to change the congestion window.