F# in the Real World
summary
Gamesys social team uses F# to run backends for 1 million DAU and about 250 million daily requests, claiming a fivefold server efficiency gain and over 90% reduction in ops cost after switching to a stateful architecture. Implementation details: F# union and record types for domain safety, async workflows and computation expressions, actor model with stateful actor-based workers backing player state to S3, optimistic versioned updates, and genetic algorithms for game tuning; concrete latency and cost numbers are given.
AI-written summary · about
chapters (19)
- 0:00Start
- 6:39closed hierarchy
- 9:11immutable by default
- 9:31lightweight syntax for types & hierarchies
- 19:13non-blocking 1/0
- 23:00x efficient improvement
- 23:1260% latency drop
- 26:19triggered by different abstraction layers
- 26:30non-functional requirements
- 27:09message-broker pattern
- 28:29need lots of facts
- 31:41auto-tuning trapping stats
- 33:00genetic algorithms
- 40:20great performance
- 42:54Disguise complexity
- 44:46Provider F# type provider for Amazon S3
- 46:38compile time validation
- 47:05no code generation
- 49:46tame complexity
related talks
52mF# in the real world
01h 08mWhy Every Day as a Developer Would Be Easier with F#
57mF# Code I Love
01h 03mF# for Performance-Critical Code
01h 00mFunctional-first programming with F#
01h 07mBuild Life-Saving Apps with Functional DSLs in F#
44mLife after business objects. Confessions of an OOP veteran
58mWhy you should use F#