4. First lets talk about Erlang
• Elixir runs on the Erlang VM
• Many of the standard lib function calls in Elixir
call Erlang functions directly.
• Elixir piggy backs on the Erlang ecosystem
5. Why Erlang VM?
• 30 year old battle tested VM
• Highly available, fault tolerant, and distributed
6. History of Erlang
• Developed by Ericsson
• Used and developed for Telecommunications
• In fact its standard library is the OTP (Open
Telecom Platform)
8. Built with Erlang
• 50 person engineering team
• managing 900 million users.
• Up to 2 million active users per server.
https://blog.whatsapp.com/196/1-million-is-so-2011
9. 2 Million Active Users?
• Intel Xeon CPU x5675 @ 3.07GHz
24 Cores - 96GB Ram
• Using only 40% of the CPU/Memory
https://blog.whatsapp.com/196/1-million-is-so-2011
10. So Erlang is pretty awesome
• Wouldn’t it be nice if we could harness this
power?
• Why isn’t everyone using Erlang?
11. Elixir
• Elixir utilizes much of the power of Erlang
• While giving us a much nicer syntax and api to
work with.
12. History of Elixir
• Created by Josè Valím - (Joe-zay Val-eem)
• First released in 2012
15. Pipe Operator
• The following code is valid elixir.
• With the pipe operator we can write our code this
way instead. Keeping us out of nested function
hell.
20. You’re already familiar with
immutability you just don’t know it.
• a = 99
• We all know 99 is always 99.
• integers are immutable values, even in Ruby.
• You’d be upset if someone monkey patched 99
to be 3 right?
21. Why is Immutability a good
thing?
• Immutability provides a contract that each piece
of information has one source of truth.
• Concurrency is a natural side effect of
immutability.
22. Does this mean I can never
change a variable once I set it?
• The short answer is no, you can reassign
variables as much as you like.
• Erlang would not be so kind, but Elixir we have
this luxury.
23. So how is it immutable if I
can change things?
• You can change things, elixir will just copy the
values you don’t change, but the original
reference stays intact.
• The original reference can then be used in part
or in whole later in the application.
24. Isn’t this inefficient?
• At first glance it would seem that way, but the
opposite is true.
• Won’t all the old reference values balloon elixirs
memory?
• Yes this means you could have thousands of
unused variables floating in memory
26. Concurrency fixes
everything
• Because the data is immutable we can rely on it
always being a certain value
• which means we can make our programs
concurrent and not worry.
• Elixir uses its concurrency to help with garbage
collections. Each process gets it own heap
allocated and that heap is garbage collected or
reclaimed as needed for that individual process.
27. Concurrency
• Elixir will eat as many cores/computers as you
can throw at it.
• It treats each core as though it were a computer
on the network with near zero latency.
32. Phoenix Framework
• It’s not rails for Elixir
• It has been influence heavily by rails as well as
many other frameworks.
• Focus on API and web sockets
33. Speed
• Web request response time is often measured in
micro seconds and not milliseconds.
34. Benchmarks
• Recently benchmarked 2 million+ web socket
connections
• With 2 - 3 second broadcast times to all 2 million
connections
41. Mix
• Mix is essentially bundler and rake combined
• Additionally mix is your test runner.
• It’s included with Elixir
42. Rails like Generators
• mix phoenix.gen.html Post posts title body:text
• mix phoenix.gen.json User users name:string
• mix phoenix.gen.model User users name:string
43. Concurrency
• Async processing with out the need for
DelayedJob or Resque/sidekiq
• We can write code that looks like this.
44. Isolated & Concurrent
• Crashes are isolated
• Data is isolated
(GC is per process, no global pauses)
• Load Balances on IO and CPU - Efficient on a
multicore
45. Garbage Collection
• No triggering massive garbage collection
• GC per process / end of process life cycle
• Ensuring top 10 percent of requests are not vastly
slower than any other request
46. Request Pipeline
• Just a series of function calls that transform
data.
• connection
|> endpoint
|> router
|> pipelines
|> controller
47. Request Pipeline
• Phoenix makes it easy to add middleware or
plugs to this pipeline. We could easily add
authentication to this.
• connection
|> endpoint
|> router
|> pipelines
|> authenticate_user
|> controller
48. Explicitness Over Magic
• Phoenix Removes a lot of the confusion and
“magic” found in rails and replaces it with
explicit calls.
• Models, Controllers and Views are singular
named by convention.