on November 15, 2010
This book provides an excellent survey of distributed programming techniques using the Ruby platform. As a software engineer who is largely unfamiliar with Ruby, but very familiar with distributed programming, I was able to leverage the book both to understand "how you do things" in Ruby, as well as to introduce myself to the (libraries, framework tools, etc.) which make distributed programming a reality. My experience in reading this book was that it had a great flow, and a very clean presentation on the subject matter. I walked away with a deeper understanding of the Ruby language itself, as well as a mapping from "strategy/design concept" to "implementation toolkit" should I have a need to write a distributed service in Ruby. In summary, the book provides an excellent survey of both distributed concepts, as well as several options available on the Ruby platform for each, covering: DRb, Rinda, RingyDingy, Starfish, Distribunaut, Politics, Starling, working with Rabbit MQ, BackgrounDRb, and Delayed Job.
Much like the Ruby language itself, the text is very concise in explaining even fairly complicated concepts. It achieves this focus of delivery by building on fundamental concepts, providing a very simple starting point, and layering on additional "would like to" one at a time, without confusing the underlying intention. This pattern is present both at the micro level, as each chapter introduces a new distributed challenge and leaves you with a working knowledge of what the Ruby space has to offer for solutions implementations; as well as at the macro level, as the topics of each chapter progress from very simple things like remote procedure calls and data marshalling, to advanced topics such as an remote work processor, distributed work queue based on a map-reduce framework or a message queue service. While the phrase "map-reduce" is mentioned in several topic headings, there is no actual example of a problem solved with a map and reduce against a distributed dataset. The final topic, although not directly a distributed programming concept, is useful information for any production system that does any work of interest: a work scheduler using BackgrounDRb.
In a similar way, each chapter presents a very tiny example app, with complete code, and a walk through, distinguishing between the concept at hand, and the library specific implementation semantics. Also, similar examples are used where appropriate, making it easier to understand the specific nuances of a particular library. Mirroring the measured build up of the text, the code samples evolve in a simple and natural way over the course of a given topic. Although one might complain that there is little imagination in the examples, this focus ensures that the reader walks away with a clear understanding of the "Hello World" level implementation - details are left up to further research.
My only complaint in terms of coverage is that the book is overly "Ruby-only" focused. As an engineer working in a largely heterogeneous environment, I would be interested in a comparison of the internal "Ruby" only packages (e.g. DRb or Rinda) vs. tools that are a bit more cross platform: e.g. in this case, a comparison against a more generalized, stack agnostic marshalling framework (e.g. Google's Protocol Buffers, Facebook's Thrift, Cisco's Etch, or Microsoft's M). In particular, this would signify where the Ruby platform offers a particular advantage for a particular kind of distributed problem. However, as the title goes, this text is focused on the Ruby specific implementations of various technologies, instead of how Ruby as a language plays (with others) in these various areas.
Overall, I would say this is an excellent resource to use as a pointer for further research. If you are developing distributed systems using Ruby, then you are likely already aware of a certain number of these libraries; the survey of alternative options might be informative. For someone like myself who is familiar with distributed systems development but relatively ignorant of Ruby, it proved to be an excellent introduction both to the language, as well as to the Ruby specific semantics for implementing a basic distributed design (e.g. marshalling, task / job execution, etc.)
(Reposted from Bay APLN [...])