10th IEEE International Conference on Network Protocols

Paris, France

November 12-15, 2002

If you have any questions or comments, please contact Leon Poutievski

updated:Nov-7-02


TUTORIALS

Tuesday, November 12

 

Morning (9:00 - 12:30)


Afternoon (14:00 - 17:30)

Technical Program

Wednesday, November  13  

 

8:30 Registration

9:15 Welcome

9:45 Paper Session 1: TCP Wired I

Chair: C. Barakat, INRIA

10:35 Pause

11:00 Keynote I

  • Architecture and Design of new Multimedia multicast Protocols
    Michel Diaz

11:45 Paper Session 2: TCP Wired II

Chair: C. Barakat, INRIA

12:35 Lunch

14:00 Paper Session 3: AD HOC/WIRELESS

Chair: I. Matta, Boston University

15:40 Pause

16:10 Paper Session 4: ROUTING

Chair: A. Cavalli, INT Evry

Thursday,   November 14

 

8:30 Paper Session 5: FORMAL METHODS

Chair: H. Ural, U. Ottawa

9:20 PANEL Retrospective -- 10 Years of Networking Research

Chair: E. Biersack, Institut Eurecom

10:30 Pause

11:00 Paper Session 6: POTPOURRI

Chair: Harmut Koenig, Brandenburg University of Technology

12:40 Lunch

14:10 Poster Session

15:30 Paper Session 7: WIRELESS TCP AND OVERLAYS

Chair: M. Buddhikot, Bell Labs

Evening Banquet: Cruise on the River Seine

 

FRIDAY,   November 15

 

9:00 Keynote II

  • Mobile Adventure with Next Generations Research
    Atsushi Murase, President & CEO DoCoMo Communications Laboratories Europe GmbH

 

9:45 Paper Session 8: MULTICAST

Chair: B. Levine, University of Massachusetts

10:35 Pause

11:00 Paper Session 9: INTERNET

Chair: T. Griffin, AT&T Labs

12:15 Lunch

13:30 Paper Session 10: SECURITY/DOS

Chair: T. Faber, ISI

14:45 Pause

15:00 Paper Session 11: PROVISIONING AND RESOURCE SETUP

Chair: J. Charzinski, Siemens

16:15 CONFERENCE ENDS

TUTORIALS

Tuesday, November 12

 

Tuesday Morning (9:00 - 12:30)


Tutorial A:  BGP
(T. Griffin, AT&T Labs)

Global Internet routing is accomplished with the Border Gateway Protocol (BGP). This protocol exchanges IP reachability information between autonomously administered networks. This tutorial presents an in-depth look at BGP, how it works, and how it is configured by ISPs. Topics covered will include

  • The economics of running an ISP
  • How economic considerations shape routing policies
  • How complex routing policies are implemented with BGP
  • The dynamic behavior of BGP
  • Recent extensions defined for BGP
  • The major challenges facing interdomain routing as the Internet continues to grow

Attendees are expected to have some familiarity with basic IP addressing and forwarding. Some understanding of routing with interior gateway protocols, such as RIP or OSPF, will be helpful but is not required.

Speaker's biography: Tim Griffin is a member of the IP Network Management and Performance Department at AT&T Labs in Florham Park, New Jersey. He received his undergraduate degree in mathematics from the University of Wisconsin, Madison, his MS and Ph.D. degrees in Computer Science from Cornell University. Before joining AT&T Labs he was a member of technical staff at Bell Laboratories. His current research interests include interdomain routing and the analysis and modeling of BGP.

Tuesday Morning (9:00 - 12:30)


Tutorial C: Internet Traffic - Characteristics, Performance and Models
(J. Charzinski, Siemens Information and Communication Networks)

This tutorial will give an overview of user and application traffic characteristics to both researchers and performance engineers interested in understanding Internet traffic, simulating Internet users or applications or dimensioning IP based networks. The audience should have a notion of probability and simple performance models (M/M/1) but the necessary background will be provided where needed.

Presenting results from the literature as well as own measurements, different aspects of user, application and backbone traffic behavior are characterized. In-depth results on application characteristics and performance measurements of Web and e-mail traffic on the Internet are presented. After giving an overview of modeling techniques, different models for Internet traffic are discussed. Quality of Service measures and their connection to different dimensioning formulas are presented and practical advice is given for handling Internet traffic models in simulations.

Outline

  1. Internet Basics
    • Why bother?
    • Traffic Engineering vs. Traffic Management
    • IP, TCP and UDP
  2. Measurement and Distributions
    • locations and methods for measurement in the Internet
    • measurement tools
    • probability distributions: examples and characteristics
    • power tails
    • distribution fitting
    • heavy tails, long-range dependence and self-similarity
    • Hurst parameter and its estimation
  3. User and Application Behavior
    • activity levels
    • separation between user and application behavior
    • TCP effects
    • flow definitions
    • traffic aggregation
  4. Application Behavior Characteristics
    • E-mail
    • file transfer
    • WWW access
    • multimedia applications
    • games.
    • Request and response size distributions
    • requests per session
    • locality
    • flow durations
    • packet size distributions
    • bit rate asymmetry
  5. User Behaviour Characteristics
    • Dial-up traffic characteristics: daily traffic
    • session duration
    • variation of mean holding time
    • interarrival times
    • Applications mixes: volume vs. time
  6. Backbone Measurements
    • Traffic mixes
    • daily and weekly patterns
    • directional differences
    • Internet reliability
    • packet loss patterns
    • delay distributions
    • flow aggregation properties
  7. Performance
    • Performance of Web and e-mail transfers on the Internet
    • delay components
    • delay distributions
    • client and server response times
    • bitrates observed within TCP connections
  8. Models
    • Model types
    • Models on different activity levels
    • User models, network traffic models
    • generators to drive TCP simulations
    • TCP models
    • Multilevel models
    • model validation
    • model parameterisation
  9. QoS Measures and Dimensioning
    • QoS measures for TCP based applications
    • delay contributions in practice
    • bottleneck and non-bottleneck dimensioning
    • multiplexing gain
    • rate envelope multiplexing
    • FBM model
    • processor sharing models
  10. Implications for Simulation
    • Long-range dependence and stationarity
    • infinite expectations
    • input parameters
    • deterministic scenarios

Speaker's biography: Joachim Charzinski received his Dipl.-Ing and Dr.-Ing degrees in electrical engineering from University of Stuttgart, Germany. >From 1992 to 1997 he was with the Institute of Communication Networks and Computer Engineering at University of Stuttgart (Prof. Kühn), working on the performance of Media Access Control schemes for ATM access networks and other teletraffic topics. Since 1997 he has been with the Information and Communications Networks group of Siemens in Munich, Germany, leading a research project on Internet traffic characterization and engineering. In 2001 he became scientific consultant for a large research project on next generation IP networks within Siemens. More information is available here

 

Tuesday Afternoon (14:00 - 17:30)


Tutorial B: Architecture, Design Principles and performance of a tier-1 backbone network
 (C. Diot, Sprint Labs)

This tutorial provides an in-depth description and analysis of the architecture, the design principles and the performance of a tier-1 IP backbone network. This tutorial is based on the experiences gained by the speakers from Sprint's Tier-1 IP backbone. It will cover only information that is available publicly; however, we will synthesize and analyze this information in a way that necessarily requires involvement in the design and the operation of a backbone network. The tutorial will give a comprehensive understanding of what the Internet is today; it will help understand the design choices, and explain the limits and the strengths of the Internet model. Quality of service will also be discussed extensively, from new services to the different approaches towards providing these services.  The tutorial will start with a brief discussion of the Internet design philosophy and the current hierarchical organization of the Internet into autonomous systems. We will describe the current architecture of an IP backbone, providing a historical perspective on various aspects such as link upgrades, evolution of Points of Presence (POP), etc. This description will be illustrated with examples from Sprint's IP backbone networks; elements of generalization to other backbones will also be discussed. Then we will discuss routing policies and practices covering issues such as intra-domain routing, BGP configurations, relationships with other ASes and traffic engineering practices. Next, we will explain the backbone design philosophy encompassing issues such as over-provisioning, QOS, fault tolerance and manageability.  In the last part of the tutorial, we will present results from an ongoing project to measure traffic on the backbone. These results shed lighton many key questions such as:

  • How can we dimension the backbone to support applications such as voice over IP?
  • What kind of end-to-end delay guarantees can we provide across the backbone?
  • How can we support multiple classes of service?
  • Do we need MPLS?
  • How can we avoid routing problems and denial of service attacks?

This tutorial also covers various approaches for monitoring and measuring Internet traffic and techniques for analyzing and interpreting its properties. We will survey existing approaches and techniques for Internet traffic measurement. This will include discussions on the types of data that can be captured (packet-level, flow-level, routing information, etc.), possible points of observation (end-hosts, routers, backbone links, etc.), common metrics (loss, delay, jitter, etc.) and the notions of active and passive monitoring. We will describe standard tools and information sources such as such as ping, traceroute, SNMP, etc.  We will conclude this section with a survey of pioneering projects in the area of network monitoring such as AT&T Netscope, Sprint IPMON, CAIDA, Keynote, etc. We will describe how measurement can help in the design and engineering of IP backbones by providing valuable input for resource provisioning, traffic engineering, DoS attack detection and improving network routing protocol operations. We will illustrate this section with a variety of tools, traces and observations from operational networks.  Finally we will cover various unsolved aspects of network measurement with focus on how completeness of data can impact the exactness of the observation, and how to infer the behavior of the network from local or partial observations.  We will conclude the tutorial with a discussion on the possible evolution of the Internet from the perspective of service convergence over IP.

 

Speaker's biography: Christophe Diot received a Ph.D. degree in Computer Science from INP Grenoble in 1991. From 1993 to 1998, he was a research scientist at INRIA Sophia Antipolis, working on new Internet architecture and protocols. Diot moved to Sprint Advanced Technology Laboratory in October 1998 to take the lead of the IP research group. His current interest is in the passive monitoring of the Sprint IP backbone in order to study IP traffic characteristics and to design new analytical models and traffic engineering solutions for pure packet networks. Diot is member of IEEE and ACM, and serves as an editor forACM/IEEE Transactions on Networking.

 

   

Tuesday Afternoon (14:00 - 17:30)

Tutorial D: Traffic Engineering Techniques
(O. Bonaventure, University of Namur)

 

Initially developed as a research network, the Internet has been optimized to provide a service where the network does its best to deliver packets to their destination. In the research Internet, connectivity was the most important issue. During the last years, we have seen a rapid growth and an increasing utilization of the Internet to carry business critical services such as e-commerce, Virtual Private Networks, Voice over IP, ...To efficiently support those services, several Internet Service Providers (ISP) rely on traffic engineering to better control the flow of IP packets. Traffic engineering encompasses several techniques that can be used to (i) shift traffic away from congested links, (ii) better distribute the traffic inside the network, (iii) quickly react to failures by directing traffic away from the faulty links or (iv) efficiently support Quality of Service (QoS) requirements. The tutorial is divided in three distinct parts.

In the first part, we discuss the techniques that can be used inside a single domain in pure IP networks. All these techniques rely on a careful tuning of the link-state intradomain routing protocols. A first set of techniques is to select the appropriate link metrics to better balance the traffic inside the network. These techniques require a good knowledge of the traffic demand and assume that this demand is stable. We then briefly discuss more dynamic techniques that have been proposed to balance the traffic when the traffic demand fluctuates. Finally, we conclude the first part of the tutorial with a discussion of the restoration time in case of failure in a pure IP network.

In the second part of the tutorial, we still consider a single domain, but in this part we show the benefits of using MultiProtocol Label Switching (MPLS) to forward the IP packets inside this network. A first technique is to statically establish Label Switched Paths (LSP) inside the network based on the known traffic demand. This technique is similar in principle to the settings of the link metrics in a pure IP network, but MPLS allows to better control the flow of packets inside the network. If the traffic demand is not stable, LSPs can be setup dynamically by a signalling protocol such as RSVP-TE based on the information distributed by the intradomain routing protocol (OSPF-TE, ISIS-TE) We describe the operations of these protocols in details. Finally, we describe the restoration capabilities of MPLS in case of failures.

In the third part of the tutorial, we describe the traffic engineering techniques that can be used to control the flow of IP packets across interdomain boundaries. We first discuss the characteristics of interdomain traffic and then show how a tuning of the configuration of the BGP routers can be used by a domain to control the flow of its incoming and outgoing traffic. We discuss the advantages and drawbacks of these techniques as well as the utilization of the BGP community attribute for traffic engineering purposes.

Speaker's biography: Olivier Bonaventure received his PhD from the University of Liege (Belgium) at the Research Unit in Networking headed by Prof. Andre Danthine. He spent one year at the Alcatel Alsthom Corporate Research Center in Antwerp before becoming professor and leading the network research group   http://www.infonet.fundp.ac.be  at the University of Namur during four years. Since October 2002, he is professor at the Catholic University of Louvain (Belgium). He has published more than twenty papers on Quality of Service (QoS), the performance of TCP/IP in ATM networks, Internet traffic characterization, BGP ... His current research interests include interdomain routing and traffic engineering, Internet traffic characterization and wireless IP networks. Olivier Bonaventure is the Belgian representative of the COST263 action on Quality of future Internet Services. He is on the editorial board of IEEE Network Magazine and received the Alcatel Bell and the Wernaers prizes from the Belgian National Fund for Scientific Research (FNRS) in 2001.