Abstracts of Tutorials
Marc Andreessen has bowed out of his tutorial due to business pressures.
Attendees who have registered for the tutorial have the choice of: (a) getting
a full refund, or (b) registering for any other tutorial (including a full
day tutorial) at NO additional charge. Registerees will receive further
information by email.
Hot topics in Networking , Raj Jain
Designing Protocols using Techniques from Distributed Systems,
Rethinking Client/Server Computing, Marc Andressen
Host-Network Interface Issues in High Performance Networks,
Designing Secure Protocols, Radia Perlman and Charles Kaufman
Multimedia Networking, Aurel Lazar
The Ethernet Renaissance: Key Protocol Enhancements, Henry Yang
Congestion Management in High Speed Networks, K.K. Ramakrishnan
Hot Topics in Networking including ATM, Multimedia and Wireless
Dr. Raj Jain, Professor, Ohio State U. (Code MF-RJ)
Return to contents
The field of networking is growing exponentially in both deployment
and development of new technologies. This makes it difficult for most
technical professionals to keep abreast of developments. This
tutorial is designed to give technical professionals an overview of
recent advances in networking. The tutorial begins with a
discussion of trends in networking and then provides an overview of
developments and technical issues in topics that are being hotly
debated in the networking community.
Raj Jain is a Professor of Computer and Information Science at the
Ohio State University. Before joining the University in April 1994,
he was a Senior Consulting Engineer at Digital Equipment Corporation
and was involved in design and analysis of distributed systems and
networking architectures. He received a Ph.D. degree from Harvard
University in 1978 and has taught graduate courses in performance
analysis at MIT. He is the Vice-Chair of ACM SIGCOMM and serves on
the editorial boards of several journals. Dr. Jain holds several
patents, and has written more than 35 papers on networking performance.
He has delivered keynote addresses at several international conferences.
He received the Computer Press Award for "The Best How- To-Book,
Systems" for his 1991 book "The Art of Computer Systems
Performance Analysis" published by Wiley, New York. He is an ACM
Lecturer, an IEEE Distinguished Visitor and a Fellow of the IEEE.
His latest book, "FDDI Handbook: High Speed Networking using Fiber and
Other Media" has been published this year by Addison-Wesley, Reading,
- Exponential Growth
- Technology Curve
- Impact of Standardization
- Trends in traffic
- Trends in Topology
- Overview & History
- Addressing vs Labels
- Protocol Structure
- Classes of Traffic
- ATM Adaptation Layers
- LAN Emulation
- Congestion Control: Rate vs Window debate
- Standard Documents
- Further Sources of Information
High Speed Networking Challenges
- Networking Success and Failures
- Key ingredients for Success
- Diseconomy of scale
- High Performance
- Key for ATM to succeed
- Five Generations of Networks
- New Optical Technologies/Devices
- Optical Networking Architectures
- Examples of Recent Testbeds
- Issues in Optical networking
- Power dissipation
- Sources for Further Information
- Video Formats
- Compression Fundamentals
- Discrete Cosine Transform
- JPEG, MPEG, H.261
- Video on packet networks
- Characteristics of Video Traffic
- Inter-scene and Intra-scene models
- High-Definition Television
- Multimedia over IP
- Multimedia Synchronization
- Sources for Further Information
- Mobile vs Wireless
- Wireless LANs
- Infrared, Radio, vs Spread Spectrum Technologies
- Wireless WANs
- IP Mobile Host Protocol
- Transport Layer Issues
- Mobile Computing Issues
- Sources for Further Information
Designing Protocols using Techniques from Distributed Algorithms
Dr. George Varghese, Associate Professor, Washington University ,
Return to contents
There appears to be a wide separation between protocols designed
for real networks and the many elegant distributed algorithms devised
by the theoretical community. Practitioners claim that theoreticians
work with toy problems, while theoreticians deplore the lack of
formal design in real protocols. This course shows that
techniques from distributed algorithms can be used systematically to
design and understand real network protocols.
George Varghese is an Associate Professor of Computer Science at
Washington University, St. Louis. His research interests are in
protocol design and distributed systems. He received his Ph.D. from
MIT. Before teaching at Washington University he was a member of the
network advanced development group at Digital Equipment Corp.
where he did research in emerging LAN technologies and routing
Techniques in distributed systems:
state machine approaches
Applications to real network problems:
spanning tree computation,
GVT computation in distributed simulation,
connection management for transport protocols,
ATM flow control,
Rethinking Client/Server Computing
Marc Andressen, Vice President of Technology and Co-founder, Netscape Communications Corporation
Return to contents
This tutorial will look at contemporary client/server information
systems on the Internet. We will show where they will be going
over the next few years. We will also present how this new model
for information publishing, database access, and interpersonal
communication is being applied to the enterprise in addition to its
popular use on the Internet. The tutorial will cover details
of practice and theory for the most popular client/server information
systems on the Internet.
Marc Andressen, 23, is Vice President of Technology and
co-founder of Netscape Communications Corporation. Marc founded the
company in April 1994 with Dr. James Clark, founder of Silicon
Graphics, Inc. As an undergraduate at the University of Illinois in
Champaign, Illinois, Marc helped to create the NCSA Mosaic research
prototype for the Internet at the university's National Center
for Supercomputing Applications (NCSA). Offering a friendly
point-and-click method for navigating the Internet and distributed
free to network users, NCSA Mosaic gained an estimated two million
users worldwide in just over one year.
The World Wide Web, both the current practice as well as a peek at
what the future will bring us based on new technologies now under
Emerging trends and technologies which will be used in securing the Web
and its data
Information access, database access and inter-personal communications
using these new technologies
Enterprise network applications versus Internet applications
Non-traditional uses of the technologies
Marc was named in 1994 as one of the top 50 people under the
age of 40 by Time Magazine, and was named "Man of the Year" by
MicroTimes Magazine. His company, Netscape Communications, was
selected as one of the "Hot Companies in 1995" by Information Week
Magazine. Marc earned his Bachelor of Science degree in Computer
Science at the University of Illinois in 1993.
Host-Network Interface Issues in High Performance Networks
Dr. Bruce Davie, Director, Internetworking Research Group, Bell Communications Research
Return to contents
It has become apparent that the interface between computers
(hosts) and the network is a potential bottleneck in achieving high
application-to-application performance. This tutorial will focus
on the issues that must be considered in the design and
implementation of high-performance host-network interfaces. A key
theme is that one must consider the entire networking subsystem
of the host, including host software and the network interface hardware.
The intended audience is graduate students, academic researchers,
and industrial researchers and developers. It is expected that
tutorial attendees would already have some familiarity with basic
networking concepts. The goal of the tutorial is to provide an
introduction to the many factors that must be considered in the
design and implementation of the networking subsystem. These
factors interact in complex ways, and only by understanding the
full scope of the problem can high end-to-end performance be
achieved. Attendees will learn how to make tradeoffs in designing
networking subsystems by analyzing performance of the various
components and of the whole system using real-world examples.
Bruce S. Davie is the Director of the Internetworking Research Group
at Bell Communications Research (Bellcore). He has worked on
performance analysis of packet-switched networks and is currently
engaged in research on host-network interface architectures and
protocols for gigabit networks. Since 1990 he has been a project
manager for the Aurora gigabit testbed. He has authored many
papers on host interface issues, designed and implemented the
"Osiris" 622 Mbit/sec ATM host interface, and was a guest editor of
the IEEE Journal of Selected Areas of Communications issue on host
interfacing. He presented a tutorial titled "Host Interfacing
for High Performance Networks" at the HPN '94 conference in Grenoble.
This section includes definition of terms and
a short review of key networking concepts. The motivation
for the tutorial is presented, including discussion of why host
interface issues are important and how the multi-faceted nature
of the problem makes it quite challenging.
Host Interface Functions
To define the problem space, this
section introduces the range of functions that must be performed
by the networking subsystem. The significance of per-byte and
per-packet functions and their relative impact on performance is
discussed, with particular focus on data movement costs. Several
key hardware/software tradeoffs are introduced.
Workstation Architecture The architecture of hosts is key to
the understanding of host interfaces. This section provides an
overview of key aspects of workstation architecture such as I/O
busses, other places to attach network interfaces, Direct Memory
Access (DMA) and Program I/O, memory subsystems and caches.
An Initial Example To provide a concrete basis for subsequent
discussion, and to show how some of the ideas presented in
previous sections can be put into practice, an example
host-network interface is presented. An important early host
interface, the Network Adapter Board (NAB), is used.
Transport Protocol Issues As the major end-to-end protocol
layer, the transport layer has a significant influence on the
networking subsystem. Various transport layer functions are
presented and the options for implementing them in hardware or
software are discussed. The importance of understanding the
relative cost of various operations (such as TCP header
processing, checksumming, and data copying) is presented.
Operating System Issues The host operating system is an
important factor affecting end-to-end performance. The costs of
interrupts and cross-domain data transfer, and techniques to
minimize them, are presented.
Predicting System Performance Having covered the major topics
needed to understand end-to-end performance, this section shows
by example how the performance of a networking subsystem can be
predicted. This section shows how workstation architecture,
operating system and transport protocol implementation, and the
interface design itself, all interact to affect end-to-end
ATM Issues The small, fixed packet size of ATM networks
provides some special challenges. This section discusses
segmentation and reassembly (SAR) issues and problems associated
with quality of service and traffic shaping. Different
approaches to the SAR problem are presented by examination of
example ATM interfaces.
Conclusions The major lessons and key architectural tradeoffs
Designing Secure Protocols
Dr. Radia Perlman, Novell, Inc. and Charlie Kaufman , Lotus Corp.
Return to contents
The international, intercorporate, information superhighway is a
scary place. There are spies anxious to steal our secrets. There
are tabloid reporters eager for a juicy scoop. There are
criminals hoping to steal goods and services. There are maladjusted
creatures that attempt to fill the emptiness of their lives by
destroying our data and sabotaging the integrity of the network itself.
Yet there is hope, through the magic of cryptography.
Cryptography, together with a properly designed protocol, allows us to
protect information from disclosure and modification.
Important properties of various types of cryptography
Classic pitfalls in authentication handshakes
- public key
- secret key
- message digest algorithms
The key distribution problem in a large network
- reflection attacks
- bucket brigade attacks
- off-line password guessing
- message splicing
Secure electronic mail
- Key Distribution Centers when secret-key-based cryptography is used
- Certification Authorities when public-keys are used
- contrast the KDC and CA approaches according to criteria such as
security, performance and ease of configuration.
- show how encryption and digital signatures are affected by mail
- describe and contrast PEM, PGP, and X.400.
- Lotus Notes
- NetSP (IBM)
- SPX (Digital)
- LAN Manager
- Windows NT
Charlie Kaufman and Radia Perlman are coauthors of the recently
published book "Network Security: Private Communication in a
Public World". They spent many years working in the network
architecture group at Digital Equipment Corporation where Charlie
designed security protocols and Radia designed routing protocols,
including the spanning tree algorithm used by bridges.
Radia Perlman is currently at Novell, Inc., designing routing,
distributed database replication, and security protocols. Her Ph.D.
thesis at MIT was on the design of a practical routing protocol
invulnerable to a denial of service attack. She is the author of
"Interconnections: Bridges and Routers".
Charlie Kaufman is currently at Iris, Inc., a subsidiary of Lotus Corp.,
where he is security architect for Lotus Notes. He is the chair of the
IETF working group designing security for the World Wide Web.
Dr. Aurel A. Lazar, Department of Electrical Engineering and Center for
Telecommunications Research, Columbia University
Return to contents
This tutorial is intended for researchers in industry and academia
involved in the study, design and implementation of scalable
multimedia networking architectures supporting interoperable
exchange mechanisms for interactive and on demand multimedia
applications with quality of service requirements.
Multimedia networks are enabled by two basic technologies:
networking and multimedia computing. We will limit ourselves to a
scenario of multimedia networks with Asynchronous Transfer Mode (ATM)
based transport, such as already seen in broadband ATM LANs,
and draw upon the xbind model for an open signalling architecture.
xbind provides an infrastructure for building open multimedia
computing platforms that support interactive multimedia
applications dealing with synchronized, time-based media on top of
a heterogeneous distributed environment with quality of service
Guaranteeing Quality of Service in multimedia networks calls for a
complete end-to-end solution. Therefore, our presentation will
begin with a reference model for network architectures called the
XRM. The XRM provides an appropriate context for discussing the
main issues arising in network control, network management and
media transport protocols as well as their interactions. The
distinctive feature of our tutorial will be a detailed discussion
of quality of service abstractions both for the networking and
for the computing platform, middleware (binding), and multimedia
service management issues. The presentation will be supported by
examples from our own implementations on various ATM platforms.
Further information about this tutorial is available
Aurel A. Lazar
is a professor of Electrical Engineering at Columbia
University. His research interests span both theoretical and
experimental studies of telecommunication networks and multimedia
systems. The theoretical research he conducted during the 1980's
pertains to the modeling, analysis and control of broadband
networks. He formulated optimal flow and admission control problems
and, by building upon the theory of point processes, derived control
laws for Markovian queueing network models in single control
and also game theoretic settings. He was the chief architect of
two experimental networks, generically called MAGNET. This work
introduced traffic classes with explicit quality of service
constraints to broadband switching and led to the concepts of
schedulable, admissible load and contract regions in real-time control
of broadband networks.
Professor Lazar is currently leading the
COMET project of the Center
for Telecommunications Research at Columbia University. The
main focus of this work is on building an open signalling
architecture that enables the rapid creation, deployment and management
of multimedia applications with quality of service requirements.
The Ethernet Renaissance: Key Protocol Enhancements
Henry Yang, Technical Director, Digital Equipment Corp.
Return to contents
This tutorial will focus on the most recent trends and developments
in Ethernet and the support of Ethernet in new technology like
ATM. We will cover both the myths and the realities in Ethernet.
We will examine the key designs and protocol enhancements in
Ethernet, including Fast Ethernet. We will review "the Capture Effect"
and how it can be solved. Given the development in ATM, we
will examine how ATM LAN Emulation works to support Ethernet.
We will review the key concepts, advantages and case study on
Switched Ethernet and Virtual Networks. We will review the
comparisons of the key contemporary high speed networks, including Fast
Ethernet, FDDI, 100BASE-VG, and ATM.
Ethernet & Fast Ethernet Overview
Latest in technology and protocol development in IEEE 802.3 and in
Henry Yang is a Senior Consulting Engineer and Technical Director
in the Network Adapter Group of Digital Equipment Corporation.
His research interests are in the issues of data link layer and
overall network design. Mr. Yang is a recognized authority in the
industry on issues relating to the data link layer. He has done
seminal work in the invention of the Ethernet, FDDI, 100 Mbit
Ethernet and other data link standards. He has worked on many
network adapters and controller/transceiver chips, for Ethernet and
FDDI, which were developed both by Digital and other vendors. He is
one of the key architects responsible for Digital's Ethernet,
Switched Ethernet Fast Ethernet, FDDI, and ATM.
Key design issues and their effects:Where do many implementations
get it wrong? How to get it right.
- What is capture effect and its impacts ? How to solve it?
- Full Duplex support
- Auto-Negotiation and configuration (speed & mode) protocol
Performance sensitivity in Ethernet
Case studies on Ethernet end-system/adapter and Hub realization
Switched Ethernet & Virtual LANs
- Subtle aspects of the Backoff algorithm
- Two-part Deference algorithm
- Why does the interpacket gap shrink as the packet propagates on
- Receive live-lock issues
Comparisons of Fast Ethernet
- Key concepts
- Key advantages & limitations
- Some case studies
Mr. Henry Yang received a B.S. degree in Electrical Engineering
with Honors from the University of Toronto in 1976 and an M.S.
degree in Computer Science from Northeastern University in 1986.He
has published many papers on data link design. He has over 28
pending and issued patents in data link and system design areas.
In 1991, he represented the U.S.A. at the 1991 Sino-American
Seminar on Computer Science and Electronic Technology held in
Beijing, China. He has taught several High Speed LAN tutorials at
the IEEE LCN conference and at InterOp.
Congestion Management in High Speed Networks
Dr. K.K. Ramakrishnan, AT&T Bell Laboratories
Return to contents
The problem of congestion management has been a particularly
active area of research for the last decade. The real problems of
managing congestion in the Internet and other connectionless
networks resulted in the development of several congestion control
and avoidance techniques, the most widely implemented of which is
Van Jacobson's Slow Start. Networks are getting faster, are more
richly interconnected and are growing in scale every day. The larger
bandwidth-delay products for these high speed networks poses
a continuing challenge for flow and congestion management
algorithms. While congestion control still attracts widespread research
interest, progress has been made in understanding the
problem. We understand more fully the behavior and effectiveness of
window-based schemes, and a new field of research has opened in the
design and analysis of rate based flow control algorithms.
Another fundamental change over the same time frame is the
development of switched networks, such as ATM. The issue of dealing
with congestion in connection oriented networks, apparently
easier, is more complicated by the need to address qualities of
service. Much of the recent work in congestion control has been in
the context of ATM flow and congestion control, appearing in
the literature and also under the auspices of the ATM Forum.
We will review the work done in the past decade in flow and
congestion control and avoidance and examine recent work in
congestion management for ATM.
Review of congestion control and avoidance in packet networks
congestion control and avoidance in connectionless networks (e.g.,
Jacobson's Slow Start, DECbit etc.)
dynamics: two-way traffic; phase effects.
Random Early Drop
Fairness & Scheduling Issues
Fair Queueing and Stochastic Fair Queueing
Issues in Non-cooperative environements
Rate Based Environments
Behavior of window based mechanisms in large BW-Delay networks.
Problems for rate based congestion control
Approaches: Netblt; Packet Pair; Hop-by-Hop Rate Control;
Dynamic Time Windows
Congestion Management in Switched Rate Based Networks
general problem for switched networks
how ATM cell switching exacerbates problem
general framework of solutions
Congestion Management in ATM Networks
rate based congestion control framework
EFCI based schemes
Explicit Rate Based schemes
hop-by-hop credit schemes
Static Credit Allocation
Dynamic Credit (Buffer) Allocation
evolution of the rate based proposal
K. K. Ramakrishnan is a Member of Technical Staff at
AT&T Bell Laboratories in Murray Hill, New Jersey. Earlier, he was
a Consulting Engineer at Digital Equipment Corporation, and was
most recently a Technical Director for High Performance Networks.
His research interests are in the performance analysis and
design of algorithms for computer networks and distributed systems
using queueing network models.
Dr. Ramakrishnan holds a B.S. in Electrical Engineering from
Bangalore University in India and an MS degree in Automation from the
Indian Institute of Science in 1978. He was with Digital from 1983
to 1994, after completing his Ph.D. in Computer Science from
the University of Maryland. He has worked and published papers in
load balancing, congestion control and avoidance, algorithms for
FDDI, distributed systems performance and issues relating to
network I/O. He has approximately 30 patents pending or issued in
these areas. Dr. Ramakrishnan participates in the Internet
Engineering Task Force and is a member of the End-End Group, as part of
the Internet Research Task Force. He is a technical editor for IEEE
Return to Sigcomm '95 home page
Last updated by S. Keshav on Thu Aug 24 18:24:11 EDT 1995