PromptWorks Talk Tuesdays: Ray Zane 9/13/16 "Elixir Processes"

Exploring concurrency with Elixir.

https://www.promptworks.com/

Transcript

1. elixir for a concurrent world
2. concurrency when two or more tasks can start, run, and complete in overlapping time periods.
3. os processes Pros They have separate sections of memory. No side effects: If VIM crashes, Chrome is unaffected. Cons Unfortunately, expensive...
4. os threads Pros Lighter than OS processes They do not block each other They run on separate CPUs, so they are truly parallel Cons Still pretty heavy weight Shared memory
5. green threads Threads that are scheduled by a virtual machine, rather than the OS. Pros They don't take up a lot of memory. Cons Green thread implementations normally cannot assign work to multiple processors. A blocking I/O operation can block all green threads.
6. so how heavy is "heavy"? A POSIX thread has a configurable stack size. On my machine, the default is 8Mb. JVM, .NET, RubyVM, and CPython all use 1Mb by default. Therefore, memory usage will add up quickly.
7. [Erlang] is a concurrent language – by that I mean that threads are part of the programming language, they do not belong to the operating system. That's really what's wrong with programming languages like Java and C++. It's threads aren't in the programming language, threads are something in the operating system – and they inherit all the problems that they have in the operating system. One of the problems is granularity of the memory management system. The memory management in the operating system protects whole pages of memory, so the smallest size that a thread can be is the smallest size of a page. That's actually too big. -- Joe Armstrong
8. goroutines Initial stack size is 2kb (as of 1.4), but then is dynamically allocated goroutines are green threads that are multiplexed as needed onto native threads, so there isn't a 1:1 mapping.
10. erlang processes
11. so... Not an OS process, but still isolated. No side effects. If a process crashes, other processes are not affected. Managed by the VM, rather than the OS. Immutable data structures sidestep shared mutable state. Processes communicate through message passing. Fault-tolerant
12. go on... Lightweight memory footprint. Dynamically allocated stack. All of this applies to Elixir as well. { _ , b i t s } = : e r l a n g . p r o c e s s _ i n f o ( s p a w n ( f n - > n i l e n d ) , : m e m o r y ) b i t s / 8 # 3 4 0 . 0
13. who cares? we are now able to easily push our systems to over 2 million tcp connections! -- WhatsApp (https://blog.whatsapp.com/196/1-million-is-so- 2011)
14. so how do processes talk?
15. Sources: https://www.youtube.com/watch?v=hlmMi0pWzJE http://blog.nindalf.com/how-goroutines-work/ http://www1.erlang.org/doc/efficiency_guide/processes.html https://speakerdeck.com/7grok/processes-threads-and-the- death-of-moores-law http://learnyousomeerlang.com/ https://www.infoq.com/presentations/erlang-software-for-a- concurrent-world