Title: An Area-Based Multicast Approach in IP and Overlay Networks

When: Thursday, July 29, 2004, 2:00-3:30 PM

Where: Room 503 IS Building

Who: Shuju Wu

Committee: Dr. Sujata Banerjee, Hewlett-Packard Laboratories, and Telecommunications Program, University of Pittsburgh; Dr. Richard R. Thompson , Telecommunications Program, University of Pittsburgh; Dr. David Tipper, Telecommunications Program, University of Pittsburgh; Dr. Panos K. Chrysanthis, Department of Computer Science, University of Pittsburgh; Dr. Katia Obraczka, Computer Engineering Department, University of California, Santa Cruz

Abstract: Multicasting is the core technology behind many group communications on the Internet such as media streaming and data dissemination and currently exists in two forms: IP multicasting and overlay multicasting. The objective of this dissertation is to seek solutions that solve the feedback implosion problem in IP multicasting and provide scalable and adaptive communications in overlay multicasting, both of which are closely related to the "shared fate" problem resulted from using a multicast tree.

The questions to be investigated in this dissertation focus on whether and how we could improve the performance in multicast environments through appropriate "performance area" management. This dissertation research will help improve both types of multicasting. More specifically, for IP multicasting, two feedback control schemes for reliable multicast protocols are proposed. They use different metrics to dynamically identify and distributively manage the areas and improve the performance. Then for overlay multicasting, the "area" concept is applied to improve application performance. Active Overlay Multicast (AOM) will be proposed and studied in this dissertation. It actively adapts to network faults by dynamically switching the performance-degrading sub-tree to the better positions; thus the application continues without experiencing bad performance when a better distribution path can be found. Along with AOM, a scalable and efficient overlay multicast tree construction mechanism will be proposed and evaluated. As will be shown by in-depth performance studies, these approaches are truly promising solutions for reliable IP multicasting and large long-lasting overlay multicasting such as content distributions and media streaming.