TUTORIAL 1 (M1)
Monday, August 31, 1998
TITLE: Integrated and Differentiated Services in the Internet
Dr. John Wroclawski
Laboratory for Computer Science
Massachusetts Institute of Technology
Cambridge, MA, USA
As the internet matures, the traditional best-effort service model is being challenged by
new applications and more stringent customer requirements. To meet this challenge, a
variety of mechanisms have been developed to provide enhanced Quality of Service control
within the IP framework. At the same time, these mechanisms must address many concerns
about scalability, efficiency, and increased complexity of the customer-provider
This tutorial will give the student a thorough introduction to the technologies being
developed to support QoS control in the Internet Protocol family. We'll begin with an
in-depth review of currently defined technologies such as RSVP and the Integrated Services
framework; we'll then move to a discussion of new developments such as Differentiated
Services, including technical underpinnings, open issues, and future research directions.
Although the tutorial will primarily focus on technical material, we will briefly touch on
the relation between these technologies and the evolution of corporate and network-wide
business and pricing models.
The tutorial is targeted at the practically oriented researcher or developer, who has a
need to understand the basic concepts, mechanisms, services and protocols for QoS control
that may be deployed in the future Internet.
Representative topics include:
* The Integrated Services Framework - Goals and Components
* RSVP Protocol Semantics and Capabilities
* Guaranteed and Controlled Load Traffic Scheduling - Semantics
* Mapping IP Integrated Services onto specific technologies -
ATM, IEEE 802 Networks, PPP.
* Taxonomy and comparison of differentiated service frameworks -
capabilities and components.
* Detailed discussion of the IETF's emerging Differentiated Service model.
* Sender, receiver, and zone-based payment strategies for network services.
John Wroclawski is a Research Scientist with the Advanced Network Architecture group at
MIT's Laboratory for Computer Science. He was an early participant in the Integrated
Services Internet project, the multi-institution research effort leading to the
development of the IETF's RSVP and Integrated Services specifications. He is an author of
several Internet RFC's in this area, a co-chair of the IETF's Integrated Services working
group (responsible for overall IP integrated services development) and of the ISSLL
working group (responsible for technology-specific
integrated services implementation standards). His broader research interests range from
network architecture to operating system design, with one present focus being the
development of differentiated service strategies for end-to-end QoS control.
Monday, August 31, 1998
TITLE: An Introduction to Internet Measurement and Modeling
Dr. Vern Paxson
Network Research Group
Lawrence Berkeley National Laboratory
University of California, Berkeley
Considering the Internet's pervasiveness, much less is known about its behavior than one
might expect. The two main reasons for the lack of knowledge are the difficulty of making
accurate, meaningful measurements and the dearth of realistic models for analyzing and
predicting its behavior. Both of these areas have seen progress in recent years.
In this tutorial we'll look at a wide range of Internet measurement and modeling issues.
We set the stage with a discussion of what makes Internet measurement exceptionally hard;
the tools available for taking raw measurements; common measurement errors to avoid; and
basic techniques from the theory of probability and stochastic processes. Next we survey
the state of the art -- what's known about today's Internet traffic and how we know it.
In the past five years, network modeling has undergone a paradigm shift away from Poisson
modeling, in which traffic correlations are assumed to be minor, and towards
"self-similar" or "fractal" modeling, in which large-scale traffic
correlations play a major role. This shift has been fundamentally driven by traffic
measurement studies. We'll discuss these studies, along with techniques for assessing the
presence and strength of fractal correlations; explanations for what causes the fractal
structure; and the implications of fractal traffic for network performance.
We finish with a case study: how to measure throughput along an Internet path, including
the numerous, subtle errors that are unfortunately easy to make.
Anyone interested in understanding how Internet traffic *really* behaves! And, more
generally, those interested in how to conduct sound network measurement and analysis.
Prerequisites: knowledge of internetworking, TCP/IP. Familiarity with basic statistics and
probability helpful, but not required.
Dr. Vern Paxson is a staff scientist with the Lawrence Berkeley National Laboratory's
Network Research Group, where his research focusses on Internet measurement. He has given
talks and taught tutorials on the subject at SIGCOMM, NANOG, IETF, ITC, and numerous other
meetings. He serves on the Internet Engineering Steering Group as one of two Area
Directors for the Transport area, and co-chairs IETF working groups on IP Performance
Metrics and TCP Implementation. He was awarded U.C. Berkeley's Sakrison Memorial Prize for
outstanding dissertation research, as well as awards for SIGCOMM and USENIX papers. Reach
him at: email@example.com.
TUTORIAL 3 (T1)
Tuesday, September 1, 1998
TITLE: Charging and Accounting for Internet Services
Massachusetts Institute of Technology Center for Technology, Policy, and Industrial
One Amherst Street, B40-235
Cambridge, MA 02139, U.S.A.
Swiss Federal Institute of Technology
Computer Engineering and Networks
CH-8092 Zurich, Switzerland
Today´s information society has a strong need for advanced communication
services and electronic contents. As the Internet becomes a ubiquitious,
global network for communications, Internet services provide a principal
means to communicate between customers, businesses, and private users.
Although solutions to account, charge, price, and bill for communication
services in traditional networks exist, the successful development and
application of these central tasks in the Internet is still missing. To
close the gap between today`s Internet communication demands and the
revenue received by Internet Service Providers from their customers, new
economic incentives and technical innovation is required. This tutorial
focuses on the defintion of the problemS of charging, accounting, and
pricing of Internet services; on basics to develop appropriate solutions;
and on the discussion of most recent economic models, technical networking
approaches, and pricing policies for providing efficient and useful
The tutorial is targeted to an academic and industrial audience. It is structured to
present basics on economic and technical issues of the Internet and develops and discusses
a number of techniques, approaches suitable, and solutions applicable to the Internet of
today and tomorrow. The audience will receive a solid overview and details related to the
problems of charging and accounting Internet services.
- Problem statement: Is there a need to charge for Internet services? economic
incentives, technical admission control, bandwidth-hungry applications, congestion
- Economic Background: Models, parameters, views.
- Technical Background: Internet protocols, problems, challenges.
- Overview of the Tutorial: Timing and keywords
- Economic models (Flat-fee, usage-based, flow-oriented, etc.)
- Responsive pricing
- Efficiency, markets etc.
- Comparison with "well-known" markets and goods (there are differences)
- Internet (Basics)
- Services (int-serv, diff-serv, guaranteed, controlled-load, best-effort, etc.
- QoS, QoS routing
- Flows and packets
- IP protocol suite
- Service guarantees
- Effective bandwidth
- Efficiency and performances
- Security questions: authorization of charging records, accounting data, etc.
- Billing and payment systems
- Comparison with ATM (service classes, call admission control, etc.
- Theoretical approaches: economic models and pricing for multiservice
networks, pricing the Internet (amongst others: e.g., Clark, Kelly,
- Simulations: (e.g., Gupta, Shenker et al.)
- Practical approaches: New Zealand, Swiss Switch approach (Best effort)
- Comparison: ATM (e.g., CA$HMAN and CANCAN)
- Further references
- Discussion of pros and cons, factors involved, suitability of different
pricing models, and efficieny (economic and technical)
- Reservation-based approach ETH Zuerich
- Evaluation environment: Crossbow platform
- Technical implementation: RSVP/SSP extensions, architecture
- Results and discussions: numbers and fugures, qualitative evaliuation
- Summary and
- Lessons learned
- Problems remaining open
Dr. Burkhard Stiller received his diploma degree in computer science and
his doctoral degree from the University of Karlsruhe, Germany in October
1990 and February 1994, respectively. From Januray 1991 until September
1995 he has been a Research Assistant at the Institute of Telematics,
University of Karlsruhe, being on leave in 1994/95 for a one-year EC
Research Fellowship at the University of Cambridge, Computer Laboratory,
England. Since November 1995 he is with the Computer Engineering and
Networks Laboratory TIK, Swiss Federal Institute of Technology ETH Zürich,
Switzerland as a Lecturer for multimedia communications and Research
Associatein the same area.
Aside from a number of project management tasks and participation in
national research projects of Germany, Switzerland, and the UK, his primary
research interests include architectures for multimedia communication
subsystems, Quality-of-Service models, charging and accounting methods,
resource reservation, transport protocols, tele-teaching, and ATM
Dr. Lee McKnight is a Lecturer in the MIT Technology and Policy Program; a
Principal Research Associate at the MIT Center for Technology, Policy and
Industrial Development; Principal Investigator of the Internet Telephony
Consortium; an Adjunct Assistant Professor of International Information and
Communication at the Fletcher School of Law and Diplomacy at Tufts
University; and President of Marengo Research, a consultancy. He received
a Ph.D. in 1989 from MIT; an M.A. from the School of Advanced International
Studies, Johns Hopkins University, in 1981; and a B.A. magna cum laude from
Tufts University in 1978.
Lee is co-editor of Internet Economics (MIT Press, 1997), a path-breaking
work that is the first to attempt to develop metrics for economic analysis
of Internet transactions; and co-author of The Gordian Knot: Political
Gridlock on the Information Highway (MIT Press, 1997), a historically
grounded vision of a regulatory and policy framework for the future. His
current research focuses on the global information economy, networked
multimedia standardization, national and international technology policy,
the convergence of the Internet and telecommunications industries, and
Internet telephony policy.
TUTORIAL 4 (T2)
Tuesday, September 1, 1998
TITLE: Designing and Building Gigabit and Terabit Internet Routers
Dr. Craig Partridge
Cambridge, MA 02140, USA
In the past few years, router technology and performance has been
transformed by a series of innovations such as switched backplanes, new
route lookup algorithms, and better distribution of functions with the
router. The result is routers that, except for the vendor logos, look and
perform radically different from the routers of five to ten years ago.
The goal of this course is to explain how these new high-speed routers
work. Because so much has changed, the course assumes that students
understand TCP/IP but have no background in router design. All the
important concepts in router architecture will be presented.
The intended audience is anyone interested in how routers are built,
including, but not limited to, students, researchers, and prospective
employees of router vendors. A basic understanding of how TCP/IP works is
useful, though not essential.
- How the
Internet Protocol Works
The basics of how IP works. IP addressing, checksum, TTL rules.
- Other Key
ICMP, ARP, SNMP, OSPF, BGP
- A Summary of
A walk through the primary standards document for routers.
- A Simple Router
A very simple PC-based example, to indicate where all the different pieces of a router's
architecture might sit and to provide an initial context for discussing how to build a
- Switch Design
A critical component of today's routers is a high-speed, parallel, switched-backplane. We
will look at various ways to design a switch and talk about which approaches appear best
for a router.
Forwarding engines are responsible for deciding where to send each packet. There are a
number of ways to build forwarding engines, ranging from ASIC-based to processor-based
Furthermore, in the past year, three new algorithms to lookup a router have been
announced, all of which offer a substantial performance improvement over previous
algorithms. We will examine how each algorithm works.
- IP Switching
and Tag Switching
IP switching and tag switching are technologies designed to bypass
the forwarding lookup completely. We will (briefly) talk about how they
A lot of a router's cost is actually in its interfaces. And there are some pitfalls in
interface design, especially in how it interfaces to the switch and how it buffers and
moves data internally.
- Quality of
Compounding the difficulty of interface design is the new expectation that interfaces will
support Quality of Service (where the definition of QoS varies by customer). Support for
QoS in interfaces is a still developing field and we will look at various ways to do it.
Routers do not just forward packets, they also are expected to filter out dangerous
packets using filters. In their richest form, these filters allow the router to act as a
firewall. Filtering technology is still evolving and we will look at some of the newest
- Routing and
Writing robust routing and control software is hard. While the instructor does not
claim to have any insights into how to make this software easier to write (indeed, his
major experience has been finding ways to make it more complex), we will look at features
of routing and control software that can make the programmers life more difficult.
- Common Design
There are some non-obvious but easy-to-make design errors in building a router.
Mistakes such as mismatches between forwarding engine and interface speeds, and getting
the error light patterns wrong on the interface board.
Issues: Striping Across Interfaces
To end the day, we'll look at two advanced topics. The first is striping packets
across interfaces. Simply striping packets isn't too hard, but if we try to stripe without
reordering packets, the problem becomes very difficult. We'll see why striping without
reordering is hard and look at one (partial) solution.
Issues: Terabit Routers
Today's routers move tens of gigabits of packet data per second. If you believe the
Internet growth curves, those routers will have to be replaced with terabit routers in two
years or so. As an end-of-day exercise, the class will apply the various ideas learned
earlier in the day to the problem of building a terabit router.
Dr. Craig Partridge is a Principal Scientist at BBN Technologies (a part of
GTE Corporation) where he does research on high-speed networking. Recently
he was the technical leader of a project (just completed) to build one of
the world's fastest Internet routers. Craig is a part-time faculty member
at Stanford and the author of the book, "Gigabit Networking."