Cost Optimization for Multiple-Session Overlay Streaming

The Hong Kong University of Science and Technology
Department of Computer Science and Engineering

Title: "Cost Optimization for Multiple-Session Overlay Streaming"

by

Mr. ZHANG Yusi

Abstract

This paper studies the problem of constructing multiple-session overlay 
live streaming trees for long-lived broadcast live streaming systems. The 
systems concerned in this paper include video sources and client proxies. 
Based on these concepts the problem is formulated as a degree-constrained, 
node weighted, multiple-session Steiner Tree problem with fanout dependent 
nodal costs, which is a generalization of the degree constrained, 
node-weighted, single-session Steiner Tree problem with fixed nodal costs. 
This paper devises an exploration into modeling multiple-tree overlay 
networks and also techniques to generalize the normal single-session 
problems into the corresponding multiple-session ones. Two heuristics are 
studied in the paper: a state-of-the-art one which simply iteratively 
applies single-session solutions and a novel one which generalizes the 
former ideas into multiple-session cases. The performances of the two 
heuristics are studied via simulation. From the results we gathered, our 
novel algorithm can achieve an outperformance of about 10% compared to the 
state-of-the-art one. It also performs better in terms of fairness among 
sessions as well.


Date            :       2 May 2013 (Thu)

Time            :       3:00pm to 4:30pm

Venue           :       3501 (lift 25-26)

Advisor         :       Dr. Gary Chan
                        Prof. James Kwok

2nd Reader      :       Dr. Jogesh Muppala