Swift on Server: Are We There Yet?
summary
A hands‑on evaluation compares server Swift (using Kitura and Vapor) against Java and Python with real benchmarks: lower request latency, throughput close to JVM, and memory footprint peaking around 800 megabytes peak (low as ~150 MB), while C++/Rust still beat it. ARC avoids GC pause‑the‑world issues and value types help safe concurrency, but force‑unwrapped optionals can crash the process, logging/structured logging and MDC are missing, many libraries (Office/image) are incomplete, SwiftPM/Xcode Linux debugging is awkward and Apple only publishes Ubuntu binaries.
AI-written summary · about
chapters (19)
- 0:00Intro
- 0:57Motivation
- 1:54Goals of This Talk
- 3:51Storage
- 5:34Load Profile
- 7:16Deployment
- 9:15No Garbage Collection
- 10:34Value Types
- 11:35Performance - JSON
- 12:58Performance - DB Query
- 13:34CPU Usage
- 13:53Errors
- 15:35Crashes
- 16:30Logging / Monitoring
- 18:25MDC
- 20:08Code Sharing
- 21:01Missing Libraries
- 22:03Tooling
- 23:56Summary
related talks
39mWhy Is Swift So Much Better On Server?
30mSwift on the Server
34mServer Side Swift
47mAn Introduction to the Swift Programming Language
41mServer-Side Swift State of the Union
35mServer(less) Swift
47mTime Traveling With Swift: Where We Started, Where We Are, and Where We’re Going
26mSwift beyond Apple: Making Software for the Physical World