This dissertation presents a traffic engineering scheme which explicitly controls net-work traffic and quality of service (QoS) in IP network core.
The current Internet is a TCP-controlled network, hence, the Internet congestion control relies mostly on TCP, and TCP with router mechanisms (such as queue management algorithms) significantly affects network performance on throughput, packet loss, delay, fairness, etc. Without any control for TCP traffic, the Internet does not provide any sort of guarantee. In this study, our key objective is to provide an effective tool with which Internet service providers (ISPs) explicitly control traffic and QoS in the Internet core.
First, a flow-controlled IP trunking scheme is proposed to control end-to-end TCP flows on a trunk basis without maintaining any per-trunk state in the IP network core and it achieves weighted max-min fairness with minimum rate guarantee among all source-destination pairs on a per-trunk basis in IP network core. In the proposed scheme, hierarchical flow control is achieved in that in the upper layer a weighted max-min flow control is implemented and acting on a per-trunk and edge-to-edge basis, and in the lower layer TCP flows belonging to each trunk share the per-trunk bandwidth allocated by the upper layer in their normal way. In addition, we present a distributed flow control algorithm which computes fair rate at each outgoing link in a weighted max-min fair sense and adjusts quickly the data transmission rate of each trunk depending on feedback rates delivered by a special control packet feeding back from the network.
Second, we also propose explicit TCP window control scheme to extend the flow-controlled IP trunking scheme to control individual TCP flows, in conjunction with buffer control, at the network edge. The proposed scheme controls explicitly TCP window size based on feedback control theory while regulating the buffer occupancy of each per-trunk queue at target value. The goal of this...