Lecture

Storage Capacity Allocation & Distributed Selfish Replication/Caching for Content Networks
18.01.2006
Speaker: Nikos Laoutaris, Post-doctoral fellow, Computer Science Department, Boston University.
Date: 18 January 2006 Time: 12:00-13:30
Location: "Mediterranean Studies" Seminar Room, FORTH, Heraklion, Crete.
Host: Panagiotis Tsakalides

Abstract:

In this talk I will present some recent research results from my work in the area of content networks. The first part of the talk will be devoted to the problem of splitting a given storage budget among the nodes of a content network that performs replication of content. A globally optimal solution to this problem is achieved by concurrently solving three inter-related sub-problems, which in the past have been studied only in isolation to each other, thus producing only partial solutions : (1) the content node placement problem, (2) the content node dimensioning problem, and (3) the object placement problem. I will show that the solution to the above capacity allocation problem calls for a solution to a multi-commodity generalization of the classical k-median problem of facility location theory. I will present a general framework for solving the multi-commodity k-median and obtaining an exact solution in polynomial time for tree networks and a constant factor approximate solution in polynomial time for general networks.

In the second part of the talk I will focus on the object placement sub-problem and examine the case in which the individual nodes act selfishly, i.e., disregard the social utility of the network (which has been the optimization objective of most previous work in the field) and instead, cater strictly to the minimization of the access cost of their individual local client population only. I will employ concepts from game theory in order to model the problem and construct an efficient distributed replication algorithm for producing Nash equilibrium object placement strategies. In the sequel, I will discuss the on-line version of the problem, in which content nodes perform caching/replacement instead of replication, and therefore are in position to adaptively track non-stationary user demands and fluid group membership (e.g., changes in the number of nodes, the available storage capacities, and the communication costs, etc.).

I will close the talk with a brief reference to some other related research results from my work.

Bio:

Nikos Laoutaris received the Ph.D. degree from the Department of Informatics and Telecommunications of the University of Athens, Greece, in 2004, for his work in the area of Content Networking. He also holds an M.Sc. degree in Telecommunications and Computer Networks (2001) and a B.Sc. degree in Computer Science (1998), both from the same department. His main research interests are in the analysis of algorithms and the performance evaluation of Internet content distribution systems (CDN, P2P, web caching) and multimedia streaming applications. He is currently a Marie Curie Outgoing International post-doctoral fellow at the Computer Science Department of Boston University.

Detailed CV and publications available at

http://cs-people.bu.edu/nlaout/

© Copyright 2007 FOUNDATION FOR RESEARCH & TECHNOLOGY - HELLAS, All rights reserved.