High Performance TCP/IP Networking Concepts, Issues, and Solutions

The world is undergoing a revolution in information and communication technology.
Not only the lives of citizens but also the networking technology are profoundly
affected by this revolution. Traditional wired networks are being replaced or complemented
by networks based on wireless, optical, satellite, and other media. TCP/IP
has emerged as the global Internet-working solution allowing communication over a
wide variety of media and networks. These new networking media and the new ways
of communication over these networks have given rise to a host of new performance
issues and concepts. To adapt and contribute effectively to such changes, engineers
and computer scientists must acquire a solid foundation and understanding of the
fundamental concepts that affect performance in TCP/IP networks.
Existing texts on TCP/IP focus on the presentation of the protocol details with
little coverage of the performance issues and concepts. These texts are good for a
first course on TCP/IP networking but do not provide sufficient material for those
advanced readers interested in acquiring in-depth knowledge of the performance
aspects of TCP/IP, especially in the emerging networking environment. To address
this need, we have written High Performance TCP/IP Networking: Concepts, Issues,
and Solutions, with a clear focus on the performance fundamentals of TCP/IP.
High Performance TCP/IP Networking: Concepts, Issues, and Solutions is a
comprehensive guide to the study of its topic. Our book provides an in-depth
coverage of (1) tools and techniques for the performance evaluation of TCP/IP
networks, (2) performance concepts and issues for running TCP/IP over wireless,
mobile, optical, and satellite networks, (3) congestion-control algorithms in hosts
and routers, and (4) high performance implementation of TCP/IP protocol stack.
This text has been created with an emphasis on fundamental concepts, such as
network measurement and simulation techniques, mathematical modeling of TCP
dynamics, and management of implementation overhead, which will continue to
guide new developments in TCP/IP. Although many specific networks, tools, and
protocols are discussed in the text, a continuous effort has been made to emphasize
the underlying performance issues and concepts
CONTRIBUTING AUTHORS
This book contains contributions from many leading experts actively working on
specific performance issues in TCP/IP networks. In addition to the two editors
(Hassan and Jain), who themselves wrote parts of the book, there are a total of
24 authors who wrote specific chapters of the book. One of the most challenging
tasks was to integrate these individual submissions into a coherent book. As part of
the integration effort, the editors have introduced a range of additional materials,
including learning objectives, review questions, hands-on projects, and case studies.
The editors maintained close liaison with the chapter authors throughout the
manuscript preparation process. The manuscript was reviewed and revised twice
to address the concerns of the reviewers. Substantial material was added in each
revision to further integrate the chapters and improve the quality of the book.
While most individual chapter authors were contacted for the revisions, the editors
themselves revised some of the chapters. The list of authors who contributed to this
book follows (chapters that were substantially revised by the editors are marked
with an asterisk).
Chapter 1 Introduction
Mahbub Hassan, University of New South Wales, Australia
Raj Jain, Ohio State University, USA
Chapter 2 TCP/IP Fundamentals*
Sanjay Jha, University of New South Wales, Australia
Chapter 3 Performance Measurement of TCP/IP Networks*
Yukio Murayama, Kurashilki University of Science and Arts, Japan
Suguru Yamaguchi, Nara Institute of Science and Technology, Japan
Chapter 4 TCP/IP Network Simulation
Mahbub Hassan, University of New South Wales, Australia
Sonia Fahmy, Purdue University, USA
Jim Wu, University of New South Wales, Australia
Abdul Aziz, University of New South Wales, Australia
Chapter 5 TCP Modeling
Sven O¨ string, University of Cambridge, United Kingdom
Harsha Sirisena, University of Canterbury, New Zealand
Chapter 6 TCP/IP Performance over Wireless Networks
George Xylomenos, Athens University of Economics and Business, Greece
George Polyzos, Athens University of Economics and Business, Greece
Petri M¨ah ¨ onen, Aachen University, Germany
Mika Saaranen, Nokia Mobile Phones, Finland
Chapter 7 TCP/IP Performance over Mobile Networks
Raghupathy Sivakumar, Georgia Institute of Technology, USA
Chapter 8 TCP/IP Performance over Optical Networks
Franco Callegati, Universita’ di Bologna, Italy
Maurizio Casoni, Universita’ di Modena and Reggio Emilia, Italy
Carla Raffaelli, Universita’ di Bologna, Italy
Chapter 9 TCP/IP Performance over Satellite Networks*
Arjan Durresi, Louisiana State University, USA
Sastri Kota, Loral Skynet, USA
Chapter 10 TCP/IP Performance over Asymmetric Networks
Venkat Padmanabhan, Microsoft Research, USA
Hari Balakrishnan, Massachusetts Institute of Technology, USA
Chapter 11 New TCP Standards and Flavors
Sonia Fahmy, Purdue University, USA
Chapter 12 Active Queue Management in TCP/IP Networks
Mohammed Atiquzzaman, University of Oklahoma, USA
Bing Zheng, New Focus, Inc., USA
Chapter 13 Software Implementation of TCP
Jeff Chase, Duke University, USA
Appendix A M/M/1 Queue
Mahbub Hassan, University of New South Wales, Australia
Raj Jain, Ohio State University, USA
Appendix B FreeBSD
Rui Zhao, University of New South Wales, Australia
Appendix C TCP Auto-Tuning
Mahbub Hassan, University of New South Wales, Australia
ORGANIZATION AND OUTLINE
The book is organized into five parts.
• Part I: Background. Part I provides an introduction to the book. It contains
two chapters. Chapter 1 provides a rationale for the book. Chapter 2 reviews
some of the key features of TCP/IP protocols used in later chapters in the
book to explain many performance issues. Chapter 2 reviews only the key
features of TCP/IP. A comprehensive treatment of TCP/IP protocol stack is
beyond the scope of the book.
• Part II: Performance Evaluation. Part II consists of Chapters 3, 4, and 5 and
provides detailed coverage of the tools and techniques for performance evaluation
of TCP/IP networks. Chapter 3 discusses the performance measurement
tools available for monitoring, analyzing, and benchmarking the performance
of TCP/IP networks. Chapter 4 introduces simulation techniques and discusses
two popular simulation tools. Chapter 5 is devoted to the mathematical
modeling of TCP congestion control algorithms.
• Part III: Performance in Emerging Networks. Chapters 6 through 10 examine
the performance concepts and issues for running TCP/IP in the emerging
networking environment. Although many of us think modems and Ethernet
when we think Internet and TCP/IP, this is no longer the reality. Yes, it is
true that nearly every home has a modem for Internet connection, and nearly
every organization has some version of the wired Ethernet connectivity to
the desktop. Many homes, however, are subscribing to Digital Subscriber
Loop (DSL) technologies for high-speed Internet connection, and many
organizations are deploying wireless LANs for flexibility. In the wide area,
too, we are witnessing new networking technologies such as mobile cellular
data networks (e.g., GPRS), high-speed optical backbones, and increasing
use of satellite links for long-distance and global coverage. Each of these
new technologies has given rise to some new concepts and issues for TCP/IP
performance. We have therefore dedicated a separate chapter to deal with
each of these technologies: Chapter 6 for wireless, Chapter 7 for mobility,
Chapter 8 for optical, Chapter 9 for satellite, and Chapter 10 for asymmetric
networks (e.g., ADSL).
• Part IV: Congestion Control. With the increase in networking complexities
and traffic dynamics, congestion-control algorithms employed at the end hosts
and in the network routers continue to evolve. The new congestion-control
algorithms in the TCP protocol resulted in many different TCP flavors (e.g.,
Tahoe, Reno, Vegas, and so on). Part IV consists of Chapters 11 and 12 and
presents an in-depth coverage of the congestion-control algorithms proposed
so far. Chapter 11 discusses various TCP flavors, and Chapter 12 examines the
new queue management schemes proposed for the network routers to combat
congestion in highly dynamic environment.
• Part V: Implementation. For emerging high-speed networks (e.g., 10 Gbps
Ethernet), the end-system implementation of TCP can become a performance
bottleneck. Part V (Chapter 13) summarizes critical performance issues for
TCP implementation in end systems and surveys solutions for improving bulk
transfer performance.
HOW TO USE THIS BOOK
The book is designed for use in a second course on networking with a prerequisite
course on introductory networking or data communications. Some of the possible
courses for which this book can be used include Advanced Computer Networks,
Advanced TCP/IP Networks, High Performance Networks, and Internet-working.
There is enough material in the book for a one-semester or one-quarter course
with 12 or 13 weeks of lecture. Depending on the background of the students, two
possible course compositions are given here.
Computer science students with limited background in mathematics and
hardware design can exclude Chapter 5 (mathematical modeling) and study Chapter
4 (Simulation) in more detail. Engineering students graduating in computer or
electrical engineering can spend fewer weeks on Chapter 4 and one extra week on
Chapter 5.
Professionals working as network engineers, R & D managers, research
scientists, and network administrators will also find this book valuable as a reference
to the most recent advances in TCP performance research.
LEARNING AIDS
There are many learning aids in this book:
• Learning Objectives. Each chapter starts with a list of learning objectives. The
learning objectives highlight the fundamental concepts (skills) students should
understand (master) as a result of reading the chapter and help them organize
their study goals. They assist instructors in pointing out lecture objectives.
• Further Reading Lists. Annotated reading lists at the end of the chapters
provide students with valuable resources for independent exploration on
specific topics of interest. These lists are particularly useful for professionals.
• Chapter Summaries. Summaries offer students a chance to review their understanding
of key concepts in the chapter before moving on.
• Review Questions. End-of-chapter review questions evaluate the degree to
which the student achieved the learning objectives and force the students to
think about the key concepts in the chapter. Answers to most of the review
questions can be found directly from the chapter; therefore, students are
forced to reread parts of the chapter to locate the answers. Such rereading is
often required to gain a clear understanding of many difficult concepts. The
instructor can use some of these questions for classroom discussions or class
tests.
• Hands-On Projects. For each chapter, a list of performance evaluation
experiments are provided for advanced students seeking to gain a deeper
understanding of some of the key concepts and solutions described in the
chapter. These experiments can be carried out on open platforms using freely
available software. The hands-on projects in this book cover a range of difficulty.
Some experiments can be completed in a few weeks using ns-2 simulation
software, without requiring any kernel-level programming.Other experiments
require modification of existing TCP/IP stacks in FreeBSD operating system
kernel. These experiments are quite challenging and can be given to students as
whole semester projects. Students attempting these experiments are expected
to have a good background in programming and operating systems. (Appendix
B provides a brief tutorial on FreeBSD for students with no prior background
in kernel programming.)
• Case Studies. A case study is introduced in Chapter 1 based on a fictitious,
but realistic organization with TCP/IP networking infrastructure. The same
case study is then used in subsequent chapters with some modifications to
introduce new performance problems. The running case study holds together
different chapters in the text, provides students a realistic context in which to
apply the concepts and techniques learned in the relevant chapters, and yields
a classroom discussion topic for the instructor.
• Figures and Illustrations. Many concepts throughout the book are explained
using illustrations. These illustrations help students understand complex performance
issues and concepts.
• Examples. Examples have been used where applicable to explain the use of
techniques learned from the text.
ACKNOWLEDGMENTS
The book would never exist without the contributions from the individual chapter
authors. We take this opportunity to thank all chapter authors for their expertise
and time and for putting up with our many requests throughout the preparation
of the manuscript. We are indebted to the anonymous reviewers for reading the
wholemanuscript or part of the earlier versions of themanuscript and making useful
comments. Their constructive suggestions significantly influenced the revisions of
the manuscript. We thank Professor Krzysztof Pawlikowski of the University of
Canterbury, New Zealand, for providing early feedback on Chapter 4. The author of
Chapter 11 thanks Tapan Karwa, Venkatesh Prabhakar, Farnaz Erfan, andMinseok
Kwon for their help with the simulation experiments in that chapter. Jim Wu, a
coauthor of Chapter 4, has been instrumental in fixing some of the problems we
faced in preparing the manuscript in LaTex. We gratefully acknowledge the support
of the entire production team at Prentice Hall. Finally, the first editor (Hassan)
would like to thank Professor Arun Sharma (previous head of school) and Professor
Paul Compton (current head of school) at the University of New South Wales for
providing a pleasant and stimulating environment in which to work