Designing Data-Intensive Applications: The Big Ideas Behind Reliable, Scalable, and Maintainable Systems

Essential reading for anyone working on distributed systems in any capacity
Designing Data-Intensive Applications really exceeded my expectations. Even if you are experienced in this area this book will re-enforce things you know (or sort of know) and bring to light new ways of thinking about solving distributed systems and data problems. It will give you a solid understanding of how to choose the right tech for different use cases.The book really pulls you in with an intro that is more high level, but mentions problems and solutions that really anyone who has worked on these types of applications have either encountered or heard mention of. The promise it makes is to take these issues such as scalability, maintainability and durability and explain how to decide on the right solutions to these issues for the problems you are solving. It does an amazing job of that throughout the book.This book covers a lot, but at the same time it knows exactly when to go deep on a subject. Right when it seems like it may be going too deep on things like how different types of databases are implemented (SSTables, B-trees, etc.) or on comparing different consensus algorithms, it is quick to point out how and why those things are important to practical real-world problems and how understanding those things is actually vital to the success of a system.Along those same lines it is excellent at circling back to concepts introduced at prior points in the book. For example the book goes into how log based storage is used for some databases as their core way of storing data and for durability in other cases. Later in the book when getting into different message/eventing systems such as Kafka and ActiveMQ things swing back to how these systems utilize log based storage in similar ways. Even if you have prior knowledge or even have worked with these technologies, how and why they work and the pros and cons of each become crystal clear and really solidified. Same can be said of it’s great explanations of things like ZooKeeper and why specific solutions like Kafka make use of it.This book is also amazing at shedding light on the fact that so little of what is out there is totally new, it attempts to go back as far as it can at times on where a certain technology’s ideas originated (back to the 1800s at some points!). Bringing in this history really gives a lot of context around the original problems that were being solved, which in turn helps understanding pros and cons. One example is the way it goes through the history of batch processing systems and HDFS. The author starts with MapReduce and relating it to tech that was developed decades before. This really clarifies how we got from batch processing systems on proprietary hardware to things like MapReduce on commodity hardware thanks in part to HDFS, eventually to stream based processing. It also does great at explaining the pros and cons of each and when one might choose one technology over the other.That’s really the theme of this book, teaching the reader how to compare and contrast different technologies for solving distributed systems and data problems. It teaches you to read between the lines on how certain technologies work so that you can identify the pros and cons early and without needing them to be spelled out by the authors of those technologies. When thinking about databases it teaches you to really consider the durability/scalability model and how things are no where near black and white between “consistent” vs “eventually consistent”, these is a ton of nuance there and it goes deep on things like single vs multi leader vs leaderless, linearizability, total order broadcast, and different consensus algorithms.I could go on forever about this book. To name a few other things it touches on to get a good idea of the breadth here: networking (and networking faults), OLAP, OLTP, 2 phase locking, graph databases, 2 phase commit, data encoding, general fault tolerance, compatibility, message passing, everything I mentioned above, and the list goes on and on and on. I recommend anyone who does any kind of work with these systems takes the time to read this book. All 600ish pages are worth reading, and it’s presented in an excellent, engaging way with real world practical examples for everything.

This book provides a high level overview but unfortunatly lacks quite a bit of detail

Se siete IT appassionati del vostro lavoro e volete capire cosa c’è sotto le cose che usate quotidianamente, questo è un libro da non perdere. Non è un manuale, non è una guida né un tutorial, ma fa fede al sottotitolo: è un “viaggio” nello scibile sulla gestione computerizzata di dati, che aiuta a comprendere al di là del marketing gli strumenti che abbiamo a disposizione.Il libro è densissimo (come dimostra un bell’indice analitico di 30 pagine su un totale di quasi 600), ricco di riferimenti (come dimostrano le folte bibliografie al termine di ogni capitolo, per lo più risorse online) ed è evidente il background accademico dell’autore. E’ un libro che richiede tempo nella lettura e comprensione – se non si saltano i dettagli, si intende… ma nel caso lasciate perdere.Una buona metà del libro riguarda la modifica concorrente di dati e i sistemi distribuiti, la parte più terrificante e affascinante, dove vengono minuziosamente spiegati i problemi che presentano e gli algoritmi che li risolvono (ad esclusione dei problemi “bizantini”). Ho trovato …”confortante” l’analisi dell’acronimo ACID :)Chiude con un’analisi di ciò che l’autore si aspetta per il futuro; molto interessante il concetto di “unbundling” dei database.

Un excellent bouquin, traite plusieurs problématiques de manière critique. Pas de solution miracle, on aura toujours des compromis. Une recommendation facile pour n’importe qui s’interesserait aux systèmes distribués.