Improved large-set data transport over the Internet using computational intelligence techniques.

摘要

This thesis proposes and investigates two different mechanisms in order to reduce the latency perceived by Internet users when downloading large-size files from the Internet. The first mechanism is the UDP-based File Transfer Protocol (UFTP). UFTP is designed for moving large-size files over WANs where traditional file transfer protocols are found to be very inefficient. It uses UDP packets to send data to avoid round trip propagation delay, a limiting factor in data transferring. A dynamic algorithm based on fuzzy logic or rough sets is designed for UFTP flow control. Experimental results presented show that UFTP, with flow control using fuzzy logic or rough sets, is up to 1.63 times faster than the conventional FTP.; The second mechanism is the Application Layer Routing Options (ALRO). Most file transfer architectures are based on the conventional client server paradigm with the Internet providing the interconnect infrastructure. In order to avoid/control network traffic congestion, the most common scenario is that the server effectively throttles back its transmission when congestion is inferred. But with the new mechanism ALRO, the server will not throttle back its transmission when congestion is inferred; instead, it will reroute the flow of packets along an alternative route constructed through a relay router or server. The proposed mechanism is implemented through UFTP. Experimental results presented show that ALRO (combined with UFTP) works as expected: attempting to balance the network load when congestion is inferred while still obtaining high-speed file transferring.