Efficient control and management for multicast services in the internet

Abstract

In recent years, emerging multicast applications involving data delivery from a single source to a large number of destinations, has imposed a lot of challenges on the existing Internet infrastructure as they usually require certain quality of service (QoS), transmission efficiency, scalability, security, etc. Thus efficient control and management architectures and mechanisms are essential for supporting those applications. In this Ph.D. thesis, we address two central problems in the design of services for supporting multicast applications. (1) How should efficient congestion control and reliability control schemes be designed on top of IP multicast services, to meet the requirements of multicast applications? (2) What kind of services is naturally demanded by multicast applications and how to design services to facilitate the deployment of multicast applications? Specifically, we investigate the interactions between congestion control and reliability control mechanisms, nominee selection mechanism for single-rate multicast congestion control, and buffer management for local loss recovery. Based on the investigations, we identify the problems and/or flaws in existing congestion control and reliability control mechanisms and propose effective solutions to the problems. Firstly, we propose a new framework to jointly perform local delivery and congestion control at delivery and control servers collocated with active routers. Next, we propose a new nominee (or “worst” receiver) selection mechanism to select the “worst” receiver accurately and scalably. Thirdly, we study the effects of the loss recovery design on congestion control. Fourthly, we formulate the caching policy design as an optimization problem, propose an OCT (Optimal Caching Time) caching policy and demonstrate that it solves the optimization problem. Finally, we identify a new direction to address the demands of multicast applications with community management services.