Skip to main content

FlatBuffers instead of JSON (or Protocol Buffers)

FlatBuffers is an efficient cross platform serialization library for C++, Java, C#, Go, Python and JavaScript (C, PHP & Ruby in progress). It was originally created at Google for game development and other performance-critical applications.

Why use FlatBuffers?

  • Access to serialized data without parsing/unpacking - What sets FlatBuffers apart is that it represents hierarchical data in a flat binary buffer in such a way that it can still be accessed directly without parsing/unpacking, while also still supporting data structure evolution (forwards/backwards compatibility).
  • Memory efficiency and speed - The only memory needed to access your data is that of the buffer. It requires 0 additional allocations (in C++, other languages may vary). FlatBuffers is also very suitable for use with mmap (or streaming), requiring only part of the buffer to be in memory. Access is close to the speed of raw struct access with only one extra indirection (a kind of vtable) to allow for format evolution and optional fields. It is aimed at projects where spending time and space (many memory allocations) to be able to access or construct serialized data is undesirable, such as in games or any other performance sensitive applications. See the benchmarks for details.
  • Flexible - Optional fields means not only do you get great forwards and backwards compatibility (increasingly important for long-lived games: don't have to update all data with each new version!). It also means you have a lot of choice in what data you write and what data you don't, and how you design data structures.
  • Tiny code footprint - Small amounts of generated code, and just a single small header as the minimum dependency, which is very easy to integrate. Again, see the benchmark section for details.
  • Strongly typed - Errors happen at compile time rather than manually having to write repetitive and error prone run-time checks. Useful code can be generated for you.
  • Convenient to use - Generated C++ code allows for terse access & construction code. Then there's optional functionality for parsing schemas and JSON-like text representations at runtime efficiently if needed (faster and more memory efficient than other JSON parsers). Java and Go code supports object-reuse. C# has efficient struct based accessors.
  • Cross platform code with no dependencies - C++ code will work with any recent gcc/clang and VS2010. Comes with build files for the tests & samples (Android .mk files, and cmake for all other platforms).
Why not use Protocol Buffers, or .. ?

Protocol Buffers is indeed relatively similar to FlatBuffers, with the primary difference being that FlatBuffers does not need a parsing/ unpacking step to a secondary representation before you can access data, often coupled with per-object memory allocation. The code is an order of magnitude bigger, too. Protocol Buffers has neither optional text import/export nor schema language features like unions.


FlatBuffers on GitHub

Improving Facebook's performance on Android with FlatBuffers
FlatBuffers is a data format that removes the need for data transformation between storage and the UI. In adopting it, we have also driven additional architectural improvements in our app like Flat Models. The mutation extensions that we built on top of FlatBuffers allow us to track server data, mutations, and local state all in a single structure, which has allowed us to simplify our data model and expose a unified API to our UI components.

In last six months, we have transitioned most of Facebook on Android to use FlatBuffers as the storage format. Some performance improvement numbers include:


  • Story load time from disk cache is reduced from 35 ms to 4 ms per story.
  • Transient memory allocations are reduced by 75 percent.
  • Cold start time is improved by 10-15 percent.
  • We have reduced storage size by 15 percent.

FlatBuffersSwift on GitHub
Infrastructure for FlatBuffers, contains a reader and a builder

Comments

Post a Comment

Most Favorite Posts

Lean prioritization matrix

  We are sorting the stream of stakeholder request and product ideas in a KPI-weighted table. The Matrix This 2x2 matrix low effort vs. high effort high value vs. low value is a nice and easy consumable visualization model. The Weights Suggested weights are Reach How many customers does the feature impact? Customers New/existing target groups Revenue Will it drive revenue either direct or via extended CLV? Acquisition Will the feature help drive new customers? Efficiency Does the feature help drive efficiency in customers’ lives – be that internal customers (colleagues), or external (paying) customers? Brand Does the feature enhance your brand awareness? The classification/quadrants Top left to bottom right Q1 Do it now! Q2 Break it down and put in prio sequence Q3 Gap filler Q4 Forget about it... for now. Andy Wicks - Mind the product
Post a Comment
Read more

Alpha Apps vs. App Unbundling

Aktuell wird viel über das Modell der "Alpha Apps" und "App Unbundling" gesprochen. Hier kurz eine Übersicht und meine 5 cents: Alpha Apps Die chinesische App WeChat geht noch weiter: Neben einem Messenger, vergleichbar mit WhatsApp, bietet sie einen Lieferdienst à la Lieferando, die Möglichkeit etwa das eigene Konto zu checken (wie sonst bei der Bank-App) und gleichzeitig die Chance etwa Promis zu folgen, wie es Twitter bietet. Solche Alpha-Apps können dadurch verschiedene Aspekte und Möglichkeiten des Internets verbinden und werden so zum idealen Zugangsportal zum Netz – so wie traditionell der Browser am Computer. Den Tod des Browsers bedeutet das aber noch lange nicht. Der Browser ist tot, es lebe der Browser! Wirtschafts Woche App Unbundling Unbundling steht für das Unterteilen von Apps oder verschiedener Funktionen in mehrere, eigenständige Applikationen. Aber nicht jede Unbundling Aktion wird positiv von Usern aufgenommen. Facebook Messenger ...
Post a Comment
Read more

Comcast - Automatic Testing of instable network connections

Comcast improves your test coverage by letting you simulate shitty network connections so you can build better systems. Testing distributed systems under hard failures like network partitions and instance termination is critical, but it's also important we test them under less catastrophic conditions because this is what they most often experience. Comcast is a tool designed to simulate common network problems like latency, bandwidth restrictions, and dropped/reordered/corrupted packets. Usage: $ comcast --device=eth0 --latency=250 --target-bw=1000 --packet-loss=10% GitHub
Post a Comment
Read more

KSCrash

Another crash reporter? Why? Because all existing solutions fall short. PLCrashReporter comes very close, but not quite: It can't handle stack overflow crashes. It doesn't fill in all fields for its Apple crash reports. It can't symbolicate on the device. It only records enough information for an Apple crash report, though there is plenty of extra useful information to be gathered! As well, each crash reporter service, though most of them use PLCrashReporter at the core, has its own format and API. KSCrash is superior for the following reasons: It catches ALL crashes. Its pluggable server reporting architecture makes it easy to adapt to any API service (it already supports Hockey and Quincy and sending via email, with more to come!). It supports symbolicating on the device. It records more information about the system and crash than any other crash reporter. It is the only crash reporter capable of creating a 100% complete Apple crash report (including thre...
Post a Comment
Read more

Server-driven UI (SDUI): Meet Zalandos AppCraft and AirBnB Lona

A short WTF: Joe Birch:  SERVER DRIVEN UI, PART 1: THE CONCEPT Zalando seems to follow the SDUI principle as well - defining a common design language and construct the screens on the backend while displaying them natively on the clients. They even go one step further; they implemented a mighty toolset to enable non-technical stakeholders to define their own native app screens Compass: Web tooling to create screens and bind data Beetroot: Backend service that combines the screen layout definition with the data Lapis/Golem: iOS/Android UI render engines Crazy cool! Good job, guys (when you do an open-source release?) To even move faster a Flutter based UI render engine implementation was great! See also AirBnB Lona SDUI approach Building a Visual Language Why Dropbox sunsetted its universal C++ mobile project and AirBnB its React Native implementation
2 comments
Read more

Swift on Android

Running Swift code on Android The biggest issue here is going to be a missing SwiftCore library. Right now Apple is shipping one for iOS, OS X and Watch OS. But that's it - and obviously they don't ship an Android version. However, not all Swift code requires the SwiftCore library, just like not all C++ code requires the STL. So as long as we use the subset of Swift that doesn't hit SwiftCore, we should be ok. Romain Goyet How to Use Apple Swift to Make an Android App Our best course of action to get something working quickly was to take Swift code and compile it into Java source code — and then take it into the Android development environment to finish building it into an installable app. So we started in Xcode, where we wrote the Swift code. Then we tested it, including running it on an iPhone or in the iPhone Simulator, and ran a custom compiler that parsed the Swift code and output Java source code (adjusting any necessary syntax, and substituting appropriate A...
Post a Comment
Read more

Android with Kotlin, iOS with Swift, Kotlin Native, flutter.io, React Native, PWA, Xamarin, Hybrid - which way to go?

Currently there are tons of frameworks how to get your business model to the user... and in the app store Full Native Android with Kotlin, iOS with Swift Deepest integration Single way to make sure that you have no lock-in effect with a framework, and you are f**ed, when Apple or Google disallows the usage of a specific technology... Two teams required 2x code PWA (Progressive Web App) Write offline- and push-capable PWA with web-technologies only Some native features might require hybrid native development and bridging (like In-App purchases, AR, ...) In best case: One web team only for website and app Maybe some native specialists for special features Kotlin Native Develop a shared framework with or without UI using Kotlin Native Additional native code will most probably be required Big Android team, small iOS specialists flutter.io (React Native | Xamarin | ... ) One codebase (flutter: Dart, React Native: JavaScript, Xamarin: C#) Additional native code ...
Post a Comment
Read more