Project Loom: Fibers and Continuations

Transcript

1. PROJECT LOOM Fibers and Continuations
2. Hello! I AM DANIEL CASTILLO Java Developer 2 https://mrpato.dev
3. Agenda • Concepts & Current State • Project Loom • Demo • Next steps • Questions 3
4. Concepts & Current State 4
5. Concepts Concurrency means different computations happening at the same time 5
6. Concepts A thread is the smallest sequence of programmed instructions that can be managed independently by a scheduler 6
7. Current State ◦ Threads have a significant footprint ◦ Are hard to create, to control and to understand ◦ Have a lot of baggage 7
8. Current State ◦ Synchronous applications are simpler but less scalable. ◦ Asynchronous applications are scalable, complex and really hard to debug and profile. 8
9. Current State: Monads 9 ◦ An abstract type used to represent computations, allows to the user to chain actions like a pipeline. ◦ Basically makes a composition of actions and then ask for the final result.
11. CODES LIKE SYNC, WORKS LIKE ASYNC 11
13. 13 The main goal is to make it possible to write code that is essentially simple, synchronous blocking code that is easy to read, to write, to profile and is scalable
14. 14 Project Loom ◦ Bringing lightweight concurrency to the JVM ◦ Lower footprint, because threads are too damn expensive! ◦ Make concurrency simple again!
15. 15 Fiber = Continuation + Scheduler
16. 16 Continuations A Continuation is a program object representing a computation that might be suspended and resume
19. 19 Fibers ◦ Fibers are lightweight or user-mode threads scheduled by the JVM not by the OS ◦ Very low memory footprint (If is it done right you can allocate millions)
20. 20 Fibers ◦ Basically a fiber wraps a task into a continuation and schedules it into a carrier thread • The continuation yields when the task needs to block • The continuation is continued when the task is ready to continue ◦ Default scheduler is a ForkJoinPool
21. 21 Fibers Carrier Threads Scheduler (ForkJoinPool)
23. 23 Cancellations ◦ Each fiber has a cancellation status and can’t be reset ◦ Fiber::cancel • Sets the cancel status • Unparks the fiber ◦ The cancellation is cooperative and can be grabbed by the task
24. 24 Deadlines ◦ Timeouts are hard to decompose ◦ But deadlines are easier to compose
25. 25 DEMO
26. Next steps 26 Fiber friendly & Resources
27. 27 Fiber Friendly ◦ java.util.concurrent locks ◦ java.util.Socket/ServerSocket ◦ java.nio.channels.SocketChannel (and friends) ◦ Thread.sleep ◦ JSSE implementations of TLS
28. 28 Resources ◦ https://openjdk.java.net/projects/loom/ ◦ https://wiki.openjdk.java.net/display/loom/Main ◦ Project Loom Update with Alan Bateman and Rickard Bäckman ◦ Ron Pressler - Loom: Bringing Lightweight Threads and Delimited Continuations to the JVM ◦ https://vorpus.org/blog/notes-on-structured-concurrency-or-go-state ment-considered-harmful/ ◦ https://vorpus.org/blog/timeouts-and-cancellation-for-humans/ ◦ http://250bpm.com/blog:71
29. 29 QUESTIONS?