Abstract

The explosion of the e-community has spurred the need for a more global communication paradigm. Today, more people than ever are using mobile ad hoc networks to fulfill their communication needs. In particular, LEO satellite access networks are gaining popularity as an alternative or a complimentary way to get connected. Being both wireless and mobile, these networks have a unique set of link errors including bit corruption, handoff and limited connectivity. Unfortunately, most transport protocols, like the standard TCP, are only designed to handle congestion-related errors common in wired networks. This inability to handle multiple kinds of errors results in severe degradation in effective throughput and energy saving, which are relevant metrics for a mobile ad hoc environment. A recent study proposed a new transport protocol for satellites called STP that addressed many of the unique problems of satellite networks. However, there was no explicit attempt to implement a differentiating error control strategy in that protocol. This thesis proposes grafting a new probing mechanism in STP to make it more responsive to the prevailing error conditions in the network. The mechanism works by investing some time and transmission effort to deduce the cause of error. However, this overhead is recouped by handsome gains in both the connection's effective throughput and its energy efficiency. This sender-only mechanism leverages many unique features of STP, making it simple, scalable and easy to deploy. The mechanism's unique semantics make it quick to start and finish, yet efficient by its ability to self-terminate if it has prematurely started. Most importantly, the mechanism also preserves the end-to-end semantics of STP. The simulation compares the STP protocol both with and witho...

Citations