16th Asian School on Computer Science

Date :December 11-12, 2005
Venue : Milton Bender Jr. Auditorium, AIT, Bangkok, Thailand
Organized and Supported By : Asian Institute of Technology (Thailand) and INRIA (France)

The Asian School of Computer Science is conducted annually by Asian Institute of Technology, Thailand and Institut National de Recherche en Informatique et en Automatique (INRIA), France. The purpose of the school is to provide short courses conducted by leading experts in the computer science fields enabling local participations from the Asia Pacific region.

The 16th Asian School on Computer Science is a two-days graduate course concentrating on the active research activities in Internet technology.

Topic: Trends in Internet technology research

Target audience: Lecturers, researchers, academic staff and graduate students

Course outline:

(1) Peer-to-Peer Internet by Keith Ross

In this course we survey contemporary research issues in peer-to-peer systems. P2P applications exploit the abundant resources at the edges of the Internet - including bandwidth, storage, and computation - to create new and exciting distributed applications.

A P2P file-sharing system allows users to pool files, search for files from the pool, and obtain a copy from the pool. Over the past five years, numerous architectures have been proposed and deployed for P2P file-sharing. There are two broad classes of architectures: structured systems, employing a distributed hash table (DHT) and unstructured, which are systems not deploying a DHT. Most of the commercial applications today are unstructured. Researchers have been very active in the design of structured systems.

The course covers trends in both practice and research. The focus will be on P2P file-sharing, but will also cover other P2P topics such as P2P VoIP. We will examine both unstructured P2P systems. We will survey many of the most popular file-sharing systems, including BitTorrent, FastTrack, Gnutella, and eDonkey. We will examine several architectures for DHTs. We will examine recent attacks to which P2P systems have been subject. We will also investigate the mathematical modeling of P2P systems.

Topics include:

Biography of Prof. Keith W. Ross
Professor Ross joined Polytechnic University as the Leonard J. Shustek Professor in Computer Science in January 2003. Before joining Polytechnic University, he was a professor for five years in the Multimedia Communications Department at Eurecom Institute in Sophia Antipolis, France. From 1985 through 1997, he was a professor in the Department of Systems Engineering at the University of Pennsylvania.

Professor Ross has worked in peer-to-peer networking, stochastic modeling, video streaming, multi-service loss networks, content distribution networks, voice over IP, optimization, queuing theory, optimal control of queues, and Markov decision processes. He is currently associate editor for IEEE/ACM Transactions on Networking and has served on numerous journal editorial boards and conference program committees.

Professor Ross is co-author (with James F. Kurose) of the popular textbook, Computer Networking: A Top-Down Approach Featuring the Internet, published by Addison-Wesley (preliminary edition in 1999, first edition in 2000, second edition in 2002, third edition 2004). The text is used by over 200 US universities each academic year, is widely used internationally, and has been translated into ten languages. Professor Ross is also the author of the research monograph, Multiservice Loss Models for Broadband Communication Networks, published by Springer in 1995.

From July 1999 to July 2001, Professor Ross took a leave of absence to found and lead Wimba, an Internet technology start-up. Wimba develops and markets Java-based asynchronous and synchronous voice-over-IP technologies, primarily for the on-line education and language learning markets. In April 2004, Wimba merged with Horizon-Live, a multimedia e-learning company based in New York City.

(2) Internet Topology Inference by Chadi Barakat

The Internet is becoming a network of huge size whose topology is difficult to monitor and control. At the same time, Internet operators and users are more and more interested in having an idea on the topology of the Internet and on the connectivity among the different networks and autonomous systems that it includes. The topology can be augmented by information on the characteristics of links composing the Internet (speed, delay, available capacity, error rate).

The information on the topology of the Internet serves for many purposes. For a network operator, it allows to optimize its network management as well as the connectivity with other operator's networks (e.g. SLA negotiation). For an Internet Service Provider, for example a Content Distribution Network (CDN) or a video-on-demand service provider, the information on the topology allows to optimize the placement of servers and the content management. The information on the topology is also necessary to optimize routing at the overlay level. For an Internet end user, knowing the topology may be of interest for choosing the best ISP to connect to, and for the optimization of end-to-end protocols as the lookup protocols in peer-to-peer networks and the server selection in case of replicated servers for some digital content.

The inference of the topology of the Internet can be done in an active way by injecting probe packets into the network, for example by using the results of traceroute to compute a tree that represents to the best the topology of the Internet. In some recent work, the active approach is used to compute the delay between the different entities of the network, and this information on the delay is then used to sort the entities based on their neighborhood.

The inference of the topology can also be done in a passive way by recording the routing messages exchanged between routers (particularly the BGP routers), then using these messages to infer the topology. More precisely, the routing messages are used to construct some logic tree that connect the different prefixes that figure in a routing table.

Both methods present some problems that are still not solved. One of these problems is the lack of precision in the obtained results. To get a better topology more information need to be collected, but this may present problems of scalability. Another problem in existing work is that it concentrates on the inference of the logical topology, which may not be the best thing to do to satisfy operators and users. Other parameters have to be considered as well as the delay, the available bandwidth, the quality of links, the load of network entities, etc.

In this course we overview the literature in this area and we present our recent results. We present techniques belonging to both the active and passive approaches. We also focus on the usefulness of considering enhanced models for Internet paths that can allow a better prediction of the performance of applications and we explain how these models can be used for a better topology inference.

Inferring Internet topology

1. Introduction: Definition and applications

2. Active approach:
2.1. Using traceroute probes
2.2. Localizing hosts using landmarks
2.3. Internet Coordinate Systems
2.4. Transparent server selection
2.5. MINC: Multicast Inference of Network Characteristics.

3. Passive approach:
3.1. using BGP routing tables
3.2. using traffic

4. Exploration of a new definition for topology that accounts for application requirements.
4.1. Motivation using Planetlab measurements.
4.2. Open issues.

Biography of Dr. Chadi Barakat
Chadi Barakat (cbarakat@sophia.inria.fr) is a permanent research scientist in the Planate research group at INRIA - Sophia Antipolis since March 2002. He got his Electrical and Electronics engineering degree from the Lebanese University of Beirut in 1997, and his master and Ph.D. degrees in Networking from the University of Nice - Sophia Antipolis in 1998 and 2001. His Ph.D. was done in the Mistral group at INRIA - Sophia Antipolis. From April 2001 to March 2002, he was with the LCA department at EPFL-Lausanne for a post-doctoral position, and from March to August 2004 he was a visiting faculty member at Intel Research Cambridge. Chadi Barakat is the general chair of PAM 2004 and WiOpt 2005 workshops, guest editor for a JSAC special issue on sampling the Internet and area editor for the CCR journal. He serves in the program committees of many international conferences as Infocom, IMC, PAM, WONS, ASWN and Globecom. His main research interests are congestion and error control in computer networks, the TCP protocol, voice over IP, wireless LANs, Internet measurement and traffic analysis, and performance evaluation of communication protocols.

Fee : The school's registration fee is USD 200*. Course materials, refreshments and lunches are included in this price. * early bird registration fee was USD 100 before 15th November 2005.

Accommodation :
The recommended hotel is the AIT Conference Center which is within the AIT Campus. The hotel rate with breakfast is as follows.

Single room : Standard USD 25, Superior USD 32.5
Double room : Standard USD 31.25, Superior USD 40

The hotel reservation can be done through the online registration.

Registration :
The online registration will be started in November 2005.

Enquiries are welcome and can be made to Wit Hmone Tin Latt (info at interlab.ait.ac.th), +66-2-524-6611.