Customer Reviews

Release It!: Design and Deploy Production-Ready Software (Pragmatic Programmers)

5 star:		(16)
4 star:		(9)
3 star:		(1)
2 star:		(0)
1 star:		(0)

 

 

The subtitle of this book might as well be Architecture and Design for the Paranoiac. The book lays out some critical aspects to creating and rolling out stable software systems. It's directed to those working in the enterprise arena and need the utmost from stability, capacity, and overall design. Nygard's definition of "enterprise" is somewhat broad in that he considers "enterprise" to be any system providing mission-critical support to a business. Regardless of how you define your particular software, I'm sure you'll find something useful in this book.

Nygard presents the book from an anti-pattern/pattern approach: he uses case studies to illustrate how critical errors in design or implementation (anti-patterns) have caused disasterous outages. He then moves on to show how application of solid design patterns could have avoided the problems. He also spends some time going in to detail on how some of the outages have happened, including brief discussions on network packet captures and decompiling third party drivers.

There are a lot of solid fundamentals in the book: dealing with exceptions at system integration points, thread synchronization, avoid rolling your own primative feature libraries such as connection pools, and make sure to test third-party libraries which play critical roles. The general approach of discussing anti-patterns followed by patterns is also a nice way of putting forth the material.

There are a lot of more complex bits covered as well, such as thinking ahead on how you'll deal with bots and crawlers, avoiding AJAX overkill, designing ahead for and using session. I also liked that Nygard talks about the importance of involving the customer in decisions on thresholds and other critical boundaries.

Despite the great content, there is a blistering flaw, IMO: A complete lack of solid implementation specifics. Nygard doesn't provide much detail at all on actual implementation of the critical patterns he espouses, nor does he point you in the direction of other sources of information. For example, the Timeout pattern is held up as a vital aspect in many parts of the book; however, there's no practical detail on actual code to implement a Timeout, and there's only one reference to a practical implementation. The Circuit Breaker pattern, central to many of Nygard's architecture assertions, doesn't have any code or a single reference to other material where you can find implementation details.

While I find that a major gap in the book, otherwise I think it's a solid addition to one's bookshelf. The writing style's very nice, his writing voice is light and concise, and the summary bullets in each section ("Remember This") are of great value. Additionally, there are plenty of references to other useful works. (Just not for the patterns I complained about above.)

This book is intended for architects, designers, and developers of software on which a business depends and whose failure costs money. The tone is informal and the style is easily read. Some architects may wish for more rigor and consider it too easily read but they might still benefit because it contains quite a bit of wisdom earned by experience.

The book discusses issues of uptime, failure, and maintainability with examples drawn from the author's experience and from other industries. Making the point from more than one point of view serves to drive it home.

This is not a programming book but the illumination of a problem is often improved by a snippet of code. The code is Java and is easily read by anyone familiar with programming. Having some familiarity with multi-threaded programming in following the explanations and their examples will make them a little easier to read but is not necessary to get the point. (Even if you truly have no knowledge of Java, looking up JDBC, JVM, EJB, JSP, J2EE, log4j, and servelet will not be much effort because not much knowledge of them is required.) The examples emphasize web applications because, I suppose, that's the environment most vulnerable to huge capacity requirements, more complex environments, more numerous causes of failure, and failures that are more visible.

The author's analysis of the problem space has two dimensions --- stability and capacity --- in which a given enterprise system can be located. The analysis also has two categories: general design and operations.

Stability and Capacity
A given coordinate, on the stability axis, for example, implies the presence and absence of features that improve and diminish stability. A feature that contributes to stability is the use of timeouts and one that detracts is the presence of many interfaces to other systems. The author identifies 8 such stability patterns that contribute to stability and 11 antipatterns that detract from it. Capacity has 4 patterns and 10 antipatterns.

General design and Operations
These two categories are less structured than those of stability and capacity.
"General design" contains advice about networking (integration with local and remote servers), security (principle of least privilege and managing passwords), availability (load balancing and clustering), and administration (QA vs production environments, configuration files, anticipating failure in start-up and shut-down, and an administrative interface).
"Operations" contains advice about recovery-oriented computing (surviving failure by restarting components, et al.), transparency (allowing a view of the system's internals), and adaptation (managing change).

The idea of patterns from software development is raised from the level of programming to the level of systems. I might have called them, for example, stability patterns and anti-stability patterns but the author's meaning is clear enough.

At the end of each pattern and antipattern section is a short and effective summary that begins with "Remember this". The design chapter has a summary and the operations chapter has two section summaries. The author clearly has in mind the reader's take-home lesson and the possibility that the book may be skimmed to reread a section. The book is cross-referenced, forward and back -- if an idea of current relevance is explained elsewhere in more detail, the page number is noted. For example, if an antipattern has a countermeasure identified by a pattern, then the relationship is noted with a page number.

Implementing some suggestions may make the QA phase of testing easier by making diagnosis and (white box) testing itself easier. If you want to design your software to be more reliable and easier to maintain after the QA phase of testing, then this book is for you.

It is the decisions that you make during development that will affect the quality of your release. Release It!: Design And Deploy Production-Ready Software was written to help you understand what causes good software to go bad, and to show you how to take control of the situation when it happens.

Release It! was written for architects, designers, and developers of enterprise class software systems. This includes applications, websites, web services, and EAI projects. According to the author, "enterprise class" means if the software stops, the company loses money.

All in all I found Release It! a very easy read. While the author comes from a Java and Unix perspective, the book's focus is generally neutral and grounded toward concepts and techniques that are portable across all platforms and technologies. Each of these case studies is based on real events, only the names (of people, companies, methods, and classes) have been change to protect the innocent. The detail of the information is very helpful, giving insight into the problems and their solutions.

My full review can be found over at Blogcritcs, but if you are looking for a better way to deliver your product, a way to avoid the traps and pitfalls, then Release It! is a good place to begin.

Once in a year, I tag a book as "book of the year", the best book I read during the year. 2007 is not over, but this my "2007 book of the year", I know that.
Frankly, I just bough this book because it's published by the "pragmatic programmers" and I trust these guys. The title is not even appealing. I knew quickly that I will discover many things.
For a long time, I wonder what to do to build up a system which is fine in production, but I didn't understand quite right what was needed (I know now that I really misunderstood the problem). The first thing that came to my mind was to make the software strong (a good thing to do by the way) ; the second thing that came to my mind was to make it really, really strong (which starts to be stupid).
Michael helps us to understand that systems fail anyway. But it should fail fast (and can often fail only partially), it must facilitate diagnosis and quick restart. And design must deal with that. But the author doesn't stay in general considerations, he points out specific patterns and antipatterns for the systems design, by means of stability and capacity. The vast majority of article tend to exposes how new technologies make the life so easy. The author revisit technologies and technical choices throught the production glasses: why AJAX should be considered with care, why we must think about pre-computed pages instead of ynamic composition in some cases, why caches is not a one-size-fits-all answer and so on. Another important point well illustrated: a system is software + hardware and the architecture must be though with physical deployment and hardware architecture in mind. Promotion of full independance of the architecture over the deployment is plain wrong. There are so many subjects tackled her, I can't speak about them all, sorry.
Michael knows his stuff, because he worked on very big e-commerce sites and his horror stories (they are numerous) are quite impressive. Maybe you don't work for a huge e-commrce site ? I don't. Believe me, the lessons learned here are still valuable, because they are emphasized by the size of the system.
This is a 334pages long book, with not so many pictures. I tend to prefer short books, simply because it takes time to read books. Here, each word count, it's not too long, for sure.
Oh, one last thing. I hate to give 5 stars. To deserve 5 stars, it must be hell of a book ! It is !!
Great job Micheal, please notify me for the next one.

Many of the texts on software engineering discuss following some methodology to produce an ideal design. Working developers quickly learn that the ideal is rarely reality and things happen once we release software out into the wild. Michael Nygard's "Release It!" picks up where these other books leave off.

Nygard talks about all the things that can and will go wrong in the finely crafted software we were sure was ready for production. A full two-thirds of the book is focused on capacity and stability issues including patterns and anti-patterns for both. The remainder of the book deals with general design issues as well as maintaining health and status in an operational system. "Release It!" provides many first hand accounts to illustrate his points, beginning with the Exception that grounded an airline, and these stories serve as excellent motivators. It's better to learn from the mistakes of others, and I really appreciated the detail Nygard went into addressing some of these horror stories.

The Pragmatic Programmers have a few "must read" books and "Release It!" is another one. After reading it and heeding its advice, you'll feel a bit better knowing that your software is better prepared for the rigors of production.

§
Going through the book, I kept flashing on a "Simpsons" episode where Homer is convinced he is going to die before morning and imparts words of advice to Bart. Bart, uncharacteristically, is all ears and remarks in surprise "This is good stuff!" Developers of enterprise-level software will certainly repeat the same words while studying this book.

A key element of interest for those same people, often too busy to read books, is that "Release It!" is organized around case studies that the author says are "taken from real events and real system failures that I have personally observed. These failures were very costly --and embarrassing-- for those involved."

Pragmatic life experience sets the tone for the book. For starters, follow the dramatic story of the failure of a software system at a major airline, then follow the analysis of what went wrong at a very concrete level. Further, abstract it to get the full value: "How do we prevent bugs in one system from affecting everything else?" That is the cycle of the book.

I don't quite see things at the higher level of the author. I work the HTML/CSS/JavaScript front ends of Java and Ruby on Rails browser-based application development. Still, my years of experience at this level affirm the statement of the author when it comes to enterprise integration: "Despite lip service, companies didn't really get off the starting line for enterprise integration until they needed to create dynamic websites. Those projects were the impetus that finally forced many companies to integrate systems that have never played well together." Amen brother.

Michael Nygard's "Release It!" is a valuable transfer of knowledge which will be appreciated by a broad layer of people in our field. As an advanced practitioner and advocate of modern W3C coding practices and Cascading Style Sheets, it was personally very gratifying to see someone with Nygard's level of view so clearly state the value of CSS and my own small part in this complex process. Thank you.
§

If you are looking for a low level book that gives you code examples and helps you implement things then you are looking in the wrong place. This book is an excellent book at high level help. The reader will have to do their own homework in the sense of application. However, I think that most programmers should read this to make themselves aware of these things. I got sick of hearing things that were unique to Java as well. But if you can get over these things then you'll like the book.

If you ever "go live" and start getting the "oh sh#$" queasy feeling, you need to read this book.
To be perfectly clear, this is not a software design patterns book in the mold of GOF, POEAA, POSA etc, though it is chock full of failure scenarios and how to handle it. This book is not about how to pass QA(zero high priority bugs & deliver all functionality), but how to design your system for the real world. When your user base multiplies by a factor of 10 and you still need to handle the 5 9s; when an external dependency fails or responds sloooooowly; when you run out of threads etc.

Buy it, read it and become a better engineer.

-Venu

This book helps you to understand why creating production ready code requires work at many levels: creating the deployment architecture, operations scenarios, and just plain writing code and testing. The book covers issues that every architect, developer and release engineer should know. The book has principles, patterns, and resources to help you identify production problems, prevent them, and survive them when all else fails.

One of the better features of the book is the stories that help you to understand issues and demonstrate that Nygard's lessons are based in experience. The lessons are (as the name of the series suggests) pragmatic, and each chapter leaves you with enough information to make changes to improve your application.

In books of this sort there is a balance between "principles" that are timeless and "how to" techniques that you can use immediately. This book is biased slightly towards Java, and contains a few references to current tools, there are enough general principles that there is little in the book that will date it.

As I read the book I found information about many topics that arose in each project that I have worked on, as well as things that I felt that I needed to learn more about.

After I finished the book I felt energized to do things better. If you build enterprise applications buy this book to learn how to build more production-ready applications. If you already know the lessons in the book, buy a copy or two for your colleagues who may not so that your life will be easier and you can get fewer late-night phone calls about a system you helped build.

As a project manager working on enterprise systems in a Fortune 500 company, I find myself coming back to this resource frequently -- maybe too frequently, as software has an amazing ability to fail in ways I don't expect. My development team has gained much from stability patterns outlined in this book, such as circuit breaker and bulkhead. I highly recommend this book as a guide to working in the real world of releasing software. -Greg