| As the title suggests, this dissertation deals with two aspects of the network transport protocol: design and performance evaluation. In part I, we present the design of the WebTP transport protocol. In part II, we analyze a particular function of any reliable transport protocol, the resequencing of packets at the receiver. In part III, we analyze TCP's congestion control algorithm.; The first objective of designing WebTP is to improve the transport layer's application-support capabilities. This is achieved by WebTP's native mode. Rather than "quantizing" the application's needs into those supported by TCP or those by UDP, WebTP allows the application to have finer-grained control over its data transmission, especially over the reliability levels and the transmission order of different data units. The second objective is to improve the transport layer's network-control capabilities, including congestion control and bandwidth allocation. This is achieved by WebTP's non-native mode. The most salient feature of this mode is that WebTP does not have to run in the end-systems, but can run in special network elements, called WebTP gateways. Important features in this mode include segment-by-segment congestion control, connection aggregation and integrated congestion control, connection transmission scheduling, automatic discovery of WebTP gateways, automatic encapsulation of TCP/UDP packets. With these, WebTP gateways can be deployed incrementally and non-invasively over an IP network, forming an expanding overlay transport network, which may be used as a layer-4 approach to provided differentiated quality of service (QoS).; Reliable transport protocols such as TCP requires packets to be accepted, i.e., delivered to the receiving application, in the order they are transmitted at the sender. Packets that arrive at the receiving host may be mis-ordered for several reasons, for instance, retransmission of dropped packets, or multi-path routing. The transport layer at the receiver is responsible to temporarily buffer out-of-order packets and to resequence all packets, as a result, delaying some of them. In part II, we present results on packet-resequencing delay and resequencing queue size. First, we model packet mis-ordering by adding an IID random propagation delay to each packet and derive simple expressions for the required buffer size and the resequencing delay. We demonstrate that these two quantities can be significant and show how they scale with the network bandwidth. (Abstract shortened by UMI.)... |
