Bringing Swift to the Cloud

Transcript

1. Bringing Swift to the Cloud Ian Partridge 7th September 2016
2. 2 Ian Partridge • Senior developer in Swift@IBM • Background in runtime systems • Java garbage collection • Runtime performance analysis • Full-stack debugging • Swift@IBM @ianpartridge @alfa
3. 3 Today • Swift’s amazing open source journey • Bringing Swift to the Cloud • Getting started with server-side Swift • The future of Swift on Linux
4. Swift’s Amazing Journey 4 Q3 2014 Q1 2016
5. Open Source Swift 5 www.swift.org www.github.com/apple/swift
6. Open, inclusive development community 6 GitHub Satellite 2016 https://youtu.be/DoQiiE5oNOU?t=9m
7. Trusted, permissive licence • Apache Software Licence v2.0 • Third most popular OSS licence • After MIT, GPLv2 • Includes patent licence requirements • Trusted by community and industry 7
8. Building on the shoulders of giants 8
9. Language evolution process 9
10. IBM Swift Sandbox • Interactive sandbox for exploring Swift • Saves your work • Supports multiple versions of Swift • Deploys your Swift code in a Linux Docker container 10 https://swiftlang.ng.bluemix.net/
11. Why Swift in the cloud? • Runtime characteristics • Performance • Memory usage • Developer experience • Fun! 11
12. Swift Performance
13. Swift Memory Usage
14. Swift Performance by Memory
15. Why does this matter? GB/hr 15
16. Apple Client Deployment Server/Cloud Deployment Application-Specific Cloud Services Client Facing App Bringing Swift to the Server Foundation Swift Swift Standard Library Core Foundation Dispatch PWQ C libs GLibc Foundation Swift Swift Standard Library Core Foundation Dispatch Darwin C libs Client-specific Libraries App Libraries Server-specific Libraries App Libraries Driving Towards Consistent Runtime across Clients/Servers Server-side Environments (Built with Foundation & Libdispatch)
17. Growing server-side ecosystem 17
18. Kitura Web Framework http://github.com/IBM-Swift/Kitura 18 • Open source web framework for Swift on macOS and Linux • Inspired by Express for Node.js • Flexible routing • Pluggable middlewares • Easy to deploy
19. Kitura Architecture 19 Web Framework Web Framework Source C Helpers & Interfaces Swift Libraries SwiftyJSON SwiftMongoDB Swift Binary Foundation Package Manager C Libraries Dispatch HttpParser HiRedis CURL PCRE2 Pluggable Components
20. Developing a Kitura web application 20 myProject ├── Package.swift ├── Sources │ └── main.swift └── Tests $ mkdir myProject && cd myProject 2. Next, initialize this project as a new Swift package 1. First, we create a new project directory $ swift package init --type executable Basic Swift package directory structure:
21. Developing a Kitura web application 21 import PackageDescription let package = Package( name: "myProject", dependencies: [ .Package(url: "https://github.com/IBM-Swift/Kitura.git", majorVersion: 0, minor: 28) ] ) 3. Add Kitura as a dependency for your project (Package.swift):
22. Developing a Kitura web application 22 Kitura.addHTTPServer(onPort: 8090, with: router) Kitura.run() let router = Router() router.get("/hello") { request, response, next in response.status(.OK).send("<h1>Hello, World!</h1>").end() } 6. Create and start an HTTPServer: 4. Add a router and a path: router.get("/hello.json") { request, response, next in response.status(.OK).send(json: JSON(["Hello": "World!"])).end() } 5. Add a JSON data route
23. Developing a Kitura web application 23 import Kitura let router = Router() router.get("/hello") { request, response, next in response.status(.OK).send("<h1>Hello, World!</h1>").end() } router.get("/hello.json") { request, response, next in response.status(.OK).send(json: JSON(["Hello": "World!"])).end() } Kitura.addHTTPServer(onPort: 8090, with: router) Kitura.run() 7. Sources/main.swift should now look like this:
24. Developing a Kitura web application 24 $ swift build 8. Compile: $ .build/debug/myFirstProject & 9. Run: 10. Open your browser: http://localhost:8090/hello http://localhost:8090/ http://localhost:8090/hello.json
25. Get started quickly with Kitura • http://github.com/IBM-Swift/Kitura-Sample • `make run` • http://github.com/IBM-Swift/Kitura-Starter-Bluemix • `cf push` • http://github.com/IBM-Swift/swift-watson-sdk 25
26. Adding a backend Apps need persistence – we can help you! • Apache Cassandra • https://github.com/IBM-Swift/Kassandra • Redis • https://github.com/IBM-Swift/Kitura-redis • CouchDB • https://github.com/IBM-Swift/Kitura-CouchDB 26
27. Adding authentication via OAuth 2.0 • HTTP Basic Auth • https://github.com/IBM-Swift/Kitura-CredentialsHTTP • Facebook • https://github.com/IBM-Swift/Kitura-CredentialsFacebook • Google • https://github.com/IBM-Swift/Kitura-CredentialsGoogle • GitHub • https://github.com/IBM-Swift/Kitura-CredentialsGitHub 27
28. 0 10 20 30 40 50 60 70 80 90 100 0 5 10 15 20 25 30 35 0.22.0 0.24.0 0.26.0 0.28.0 0.28.1 0.29.0 CPU (% usage) Throughput (k req/sec) Throughput and CPU - macOS Throughput CPU Kitura performance - macOS 28 Steady improvements – as well as improvement attributed to the runtime 1. HTTP keepalive implementation, removal of unnecessary closures 2. 06-20 to 07-25 snapshot migration 3. Avoid reallocation of dictionaries for HTTP Headers 4. Faster HTTP date construction, Replace some internal uses of Data with UnsafeRawPointer + length offsets, Use multiple Dispatch queues for socket I/O Hardware: MacBook Pro mid-2015, 2.5ghz Core i7 Benchmark: simple ‘Hello World’ plaintext response 1 2 3 4
29. 40 50 60 70 80 90 100 0 10 20 30 40 50 60 0.10.0 0.12.0 0.14.0 0.16.0 0.20.0 0.22.0 0.24.0 0.26.0 0.28.0 0.28.1 0.29.0 CPU (% usage) Throughput (k req/sec) Throughput and CPU - Linux Throughput CPU Kitura performance - Linux 29 Steady improvements, but also some challenges 1. Reduce network fragmentation 2. Avoid self-referencing closures 3. Avoid CFGetConverter bottleneck 4. Loss of whole-module-optimization 5. HTTP keepalive implementation 6. 06-20 to 07-25 snapshot migration 7. Avoid reallocation of dictionaries for HTTP Headers 8. Faster HTTP date construction, Replace some internal uses of Data with UnsafeRawPointer + length offsets, Use multiple Dispatch queues for socket I/O Hardware: Lenovo x3550 M5, Haswell 2.6GHz (using 4 cores) Benchmark: simple ‘Hello World’ plaintext response 1 2 3 4 5 6 7 8
30. BluePic demo app https://github.com/IBM-Swift/BluePic 30
31. TodoList demo app • A shared example to showcase backend tech stacks • http://todobackend.com/ • An example using Kitura to develop a Todo-Backend • https://github.com/IBM-Swift/Kitura-TodoList 31
32. IBM Cloud Tools for Swift • Deployment made easy • Clone, code, push • Demo projects to try 32 http://cloudtools.bluemix.net
33. IBM Swift Package Catalog 33 http://swiftpkgs.ng.bluemix.net • Create, share and discover new Swift packages • Explore dependencies • Check licenses • Star your favourite packages
34. The future of Kitura? • Next-gen transport protocols • WebSockets / socket.io • HTTP/2 • TLS • Backend abstraction through ORM • More performance • Monitoring • Clustering 34
35. The future of Swift on Linux? • Completing the implementation of Foundation • Libdispatch performance and scalability improvements • Building a vibrant Swift on Linux community • Server-side packages in the rapidly-growing SPM ecosystem • IDE support • More cloud integration • Exploring the Swift “sweet-spot” • Linux-driven enhancements to Swift itself? 35
36. Find out more 36 https://developer.ibm.com/swift/ The Swift@IBM devCenter Join IBM Swift Engineering and learn the latest news