This book is all about wrangling a herd of network computers so that all
display the correct time. This may seem like a really narrow business,
but the issues go far beyond winding the clock on your display taskbar.
Carefully coordinated, reliable, and accurate time is vital for traffic control
in the air and on the ground, buying and selling things, and TV network
programming. Even worse, ill-gotten time might cause domain name system
(DNS) caches to expire and the entire Internet to implode on the root
servers, which was considered a serious threat on the eve of the millennium
in 1999. Critical data files might expire before they are created, and
an electronic message might arrive before it was sent. Reliable and accurate
computer time is necessary for any real-time distributed computer
application, which is what much of our public infrastructure has become.
This book speaks to the technological infrastructure of time dissemination,
distribution, and synchronization, specifically the architecture, protocols,
and algorithms of the Network Time Protocol (NTP). NTP has been active
in one form or another for over almost three decades on the public Internet
and numerous private networks on the nether side of firewalls. Just about
everything today that can be connected to a network wire has support for
NTP: print servers, WI-FI access points, routers and printers of every stripe,
and even battery backup systems. NTP subnets are in space, on the seabed,
onboard warships, and on every continent, including Antarctica. NTP comes
with most flavors of Windows as well as all flavors of Unix. About 25 million
clients implode on the NTP time servers at the National Institute of Standards
and Technology (NIST) alone (J. Levine, NIST, personal communication).
This book is designed primarily as a reference book but is suitable for a
specialized university course at the senior and graduate levels in both computer
engineering and computer science departments. Some chapters may
go down more easily for an electrical engineer, especially those dealing with
mathematical concepts, others more easily for a computer scientist, especially
those dealing with computing theory, but each will learn from the other.
There are things for mathematicians, cryptographers, and spacefarers, even
something for historians.
Welcome to the second edition of this book. The original 16 chapters of
the first edition remain, but some have been rewritten, updated, and new
Downloaded by [Universiti Sains Malaysia (USM)] at 21:18 01 October 2017
xxiv Preface
material added. Four new chapters have been added, two of which discuss
timekeeping in space missions. The presentation begins in Chapter 1 with a
general overview of the architecture, protocols, and algorithms for computer
network timekeeping. This includes how time flows from national time standards
via radio, satellite, and telephone modem to hundreds of primary time
servers, then via NTP subnets to millions of secondary servers and clients at
increasing stratum levels. Chapter 2 describes the principal components of
an NTP client and how it works with redundant servers and diverse network
paths. Chapter 3 contains an in-depth description of the critical algorithms
so important for the consistency, accuracy, and reliability that any good computer
scientist will relish. The actual algorithm used to adjust the computer
clock is so special that Chapter 4 is completely dedicated to its description
and operation. As the word network is prominent in the title of this book,
Chapter 5 presents an overview of the engineering principles guiding network
configuration and resource discovery.
Along about now, you should ask how well the contraption works. Chapter
6 evaluates the performance of typical NTP subnets with respect to network
delay variations and clock frequency errors. It shows the results of a survey
of NTP servers and clients to determine typical time and frequency error
distributions. It then analyzes typical NTP configurations to determine such
things as processor and network overhead and engineered defenses against
flood attacks.
An NTP subnet ultimately depends on national and international means to
disseminate standard time to the general population, including Internet computers.
Chapter 7 describes a number of systems and drivers for current radio,
satellites, and telephone modem dissemination means. Chapter 8 describes
specialized kernel software used in some computer systems to improve timekeeping
accuracy and precision, ultimately to the order of nanoseconds.
In modern experience we have learned that computer security is a very
serious business, and timekeeping networks are not exempt. What may be
different for NTP subnets is that, by their very nature, the data exchanged are
public values transmitted from public servers over public networks, so servers
and clients of public networks might be seen as very inviting targets for
tempo-terrorists. In addition, there are devilishly intricate issues when dated
material such as cryptographic certificates must be verified by the protocol
that uses them. Chapter 9 describes the NTP security model and authentication
protocol, which shares headers with NTP, while Chapter 10 describes a
number of cryptographic algorithms designed to prove industrial-strength
group membership.
Computer network timekeeping, like many other physical systems, is
not without errors, both deterministic and stochastic. Chapter 11 contains
an intricate analysis of errors inherent in reading the system clock and disciplining
its time and frequency relative to the clock in another computer.
Chapter 12 is on modeling and analysis of the computer clock, together with
a mathematical description of its characteristics.
Computer Network Time Synchronization The Network Time Protocol on Earth and in Space, Second Edition