Customer Reviews

The Design of the UNIX Operating System [Prentice-Hall Software Series]

5 star:		(23)
4 star:		(1)
3 star:		(1)
2 star:		(0)
1 star:		(1)

 

 

UNIX in its many derivatives is a 30-year old operating system. Why has it stood the test of time? Because it was designed with a simple philosophy: to give the user the ability to create his own tools to solve problems. This book has been a classic in UNIX architecture since it was published in 1987. Although some of the sections are a bit dated (e.g., IPC, file systems), it is still a valuable source of information. You will begin to understand the design philosophy of UNIX after reading this book. You will see why some of the design decisions were made (primarily due to the hardware of the time) and some of the really neat kluges that was devised to get around those problems. This book deserves its status as one of classics of UNIX literature. I highly recommend it.

After becoming a Linux enthusiast, I bought the book "Understanding the Linux Kernel", which I unfortunately found to be almost completely impenetrable. I had heard good things about this book ("The Design of the UNIX Operating System") - in fact, I read somewhere that this was the book that taught Linus to write OS code. I put down the Linux book and picked up Maurice Bachs, which I found perfectly readable - it's abstract treatment of kernel algorithms made it easy to get the "big picture". Now, with Bach's complete treatment of the design philosophy under my belt, I'm going back to the linux kernel book, and I've found it to be a breeze... even the linux kernel source code itself now makes sense. This book is excellent for anybody who's serious about programming.

Maurice Bach's The Design of the Unix Operating System still holds the place of honor on my technical reference bookshelf. After almost 20 years, it provides a clear overview of basic Unix organization and operations and is a model for how technical books should be written. Readers who complain that the text is dated evidently did not bother to notice the 1986 copyright date. Its age, however, has not diminished its clarity of content or usefulness in understanding the Unix operating system. Bach deserves an award for excellence in technical writing.

While there may be more detail to be found in "The Magic Garden," or more up-to-date coverage in the likes of Vahalia or Schimmel, Bach's opus is, in the view of this twenty-plus-year UNIX guru, unmatched. I say this because only while reading Bach's book do I experience the sense of philsophic structural perfection, of tool-orientation, of practicality-versus-theoretic-efficiency tradeoff, that characterizes the earliest UNIX monographs (Ritchie, Kernighan, Bourne, Lycklama, Ossana; that sort of thing) that busied me as a freshman. Bach imparts to the reader a glorious--and gloriously holistic--depiction of the structure of the UNIX kernel as a unit. Algorithmic details are provided where appropriate. Exceptionally well thought-out exercises stimulate the reader to extend the textual material where meet. The material is assuredly out of date, but I dare you to critize, say, Lions as being "out of date" (whether or not it describes a 25-year-old, 9K-LOC kernel, it is a scripture of paramount importance, a cornerstone of my computer engineering [n.b.: I didn't say "computer science"] library).

For those who are wont to compare Leffler and Bach--if for no other reason than that they are coevals--I heartily endorse Bach over its competitor. It's nice. It's clean. It's precise. You just couldn't ask for more. And, BTW, stay away from "The Magic Garden." I'm not sure that five hundred pages worth of out-of-context code excerpts, inundating the reader with thousands of kernel variables, accomplishes much by way of imparting conceptual understanding.

(I'm reminded: a customer of mine--an older gentleman with a Ph.D. in physics--once asked me for a concise description of the workings of UNIX, something that introduced the basic concepts at a scholarly but not overweight level. I told him I had a recommendation in mind. "You're going to give me 'The Magic Garden'," John complained; "Don't bother. It stinks!" Was John ever surprised when I pointed him to the third entry in Tanenbaum's Modern Operating Systems series. It has concise thirty or forty-page entries on UNIX, MS-DOS, and a handful of others. For those who want to know--from a scientist's viewpoint--what the fundamentals of the UNIX OS and superjacent environment are, what it can do, how one navigates within it, etc., at a _conceptual_ level that trucks not with the details of Bach or Leffler, seek ye Tanenbaum II.)

This book has been around a while, but it's still hard to beat. It describes the design of the UNIX scheduler and file system. Most of the frequently used system calls are covered. The implementation the author covers is AT&T UNIX System V Release 2 and 3. The author is a fine writer. His explanations are clear, and pseudo-code is provided for most of the algorithms. Recommended for intermediate-to-advanced programmers who want to understand what's going on "under the hood."

This book is a bit dated but IMHO it is simply the best book on computer technology or computer science that I have ever read. It is not too abstract and explains how something actually works, yet it does not get lost in details. It concisely explains the fundamental data structures, algorithms, and ideas of a real operating system - one that has had a major impact on all major operating systems today.

I'm something of an OS freak (not an expert though) and I collect OS books. I've read many of the classics of the field but I think this book is the crowning achievement of OS literature. Here are the arguments to support my claim:

a) It does not go into explaining general OS theory, thus all space can be dedicated to explaining the details of one operating system (Unix System V Release 2). This of course makes it unsuitable for begginers as it assumes you have a good understanding of basic concepts like race conditions, mutual exclusion, data structures, etc. If you're a begginer don't buy this book yet; get "Operating Systems - Design and Implementation" by Tanenbaum & Woodhull or "Operating System Concepts" by Silberschatz, Galvin and Baer.

b) It details EVERY algorithm with C-like pseudocode and adds verbal explanations exemplifying operations running through the algorithms. This is unlike other OS books which sometimes just give general descriptions of algorithms with no examples.

c) Explanations are complemented by many diagrams of data structures in various states of manipulation by the algorithms. This is possibly the most valuable feature of the book as it does wonders to help you understand what the kernel is doing; you get to 'see' how the algorithms work. This sets it apart from practically all other OS books I've read that just mention in passing "... then function 'x' manipulates data structure 'y'" and leave you to find out the implications of these manipulations. Diagrams also make the book superior to mere code listings.

d) Each chapter 'uses' the algorithms explained in the previous chapter to explain higher level functionality. This is much unlike other OS books which are just unstructured and make you loose the big picture of how the various pieces fit together. Chapters also start with an introductory overall view of the current topic.

So, what is not to like about this book? The only thing I can think of is that it deals with a 'dead' OS. Unix System V only runs in a handful of computer installations these days (if any), while its derivatives have changed too much to serve as a reference while reading the book. Still, System V binaries and source are available on the internet, legally of course. Search for The Unix Heritage Society archives. If you want to get really hardcore you can even get a PDP-11 emulator and set up Sys V in it. There are, of course, other books that delve into present day operating systems; "Solaris Internals" , for instance.

Also, Unix-haters might point out this is just another book on Unix. Well, unfortunately there are no books that explain, say MS Windows, at this level of detail; blaim MS. But still, while dealing with the specifics of one single OS, you do get a general understanding of how other OS's might work.

In my humble opinion this book is the 'King of the Hill' of OS literature; it has helped me finally understand things like context switching and memory mapping. An absolute feast to read, particularly if you like Unix.

What can I say that hasn't been said already? Still, I have to say that this is a great book on the UNIX operating system. It explains all the kernel algorithms in meticulous detail. It is required reading for any computer science students and especially students taking an operating systems class.

This book starts of with a overall good introduction of the unix process and the file system and then goes on building it from the basics (the os) with well written algorithms which are highly modular and help in clear cut understanding of the operating system.It finishes with the extension of these algorithms to multiprocessor systems.This is a very fine and well written book and i would recommend it for anyone who is interested in the the various issues in the design of an operating system and wants to design one

This is an excellent text for students of the 'Operating Systems' course, and deals with the innards of UNIX. Any UNIX enthusiast will love this book. This book is not for beginners as it deals with the inner workings and principles involved in the design of the OS. Difficult concepts like the kernel algorithms are explained in the best possible way. A tough yet satisfying read.