data intensive applications martin kleppmann

As software engineers, the applications that we build principally consists of two things: state and application code.“This book should be required reading for software engineers…” - Kevin Scott I concur with Kevin Scott’s glowing endorsement of the book:.Mark out what catches your eye and focus on those things.Best way to let this book hold your attention is to start with the Index and the Summary.

This book covers many topics, some of which may not be immediately useful to you at the moment (eg.Beautiful illustrations at the start of every chapter.Plenty of cross-references (note: a boatload of them!) for you to learn (even) more about data systems.Effective presentation of pertinent concepts about distributed and concurrent systems.Solid suggestions on how to make scalable data systems ✅ (especially Chapter 12, Unbundling Databases, and Aiming For Correctness🎖🙆‍♂️🤯).when someone messages you, data flows asynchronously from the other person’s phone to the server, and then you receive the message via a socket) Start to think of systems in a dataflow paradigm (eg.Systems are composed of state and application, ie.Determine the required correctness of the system.Same business event should only occur at most once per day? Determine the constraints of the system.Increase system throughput by increasing partitions, instead of using coordination (eg.Consumer that sends out email should only send at most one email to the customer) the side effect of an action is guaranteed to be executed only once, even if you repeat the execution of the given action) (eg. Design system actions to be idempotent (ie.Kafka) for their durability and atomicity guarantees Use systems that rely on write-ahead-log ( WAL) (eg.

General principles of building ‘unbundled’ systems.

‘Unbundling the database’ means that you’ll have to build all the system guarantees that comes with ‘bundled’ databases by yourself.

You can ‘ unbundle the database’ in order to get more performance out of systems, but there are caveats.

transactions, 2PL, indexes, joins, foreign key constraints, MVCC, etc) “ In a massively distributed and concurrent systems scenario, how can one use databases (SQL/NoSQL) in a performant manner and yet still preserve data integrity/correctness?” Key points: Designing Data-Intensive Applications by Martin Kleppmann ( link)

Data intensive applications martin kleppmann