Enable high coucurrent & availability with Goliath

Enable high concurrent &
availability with Goliath
Von Stark
Saturday, December 8, 12

View Slide

Who
I use Ruby, Neo4j, Riak
co-founder of nosql.org.tw
co-founder at NeoArk
@vonstark32
http://vonstark.co
http://neo4j.tw
http://nosql.org.tw

View Slide

Background
Ruby Mainly
iPhone & Android Clients
Requests : data analyze, data process, graph traversing ...
Heavy I/O requests

View Slide

What we use before?
Rails + Thin / Rainbows / Unicorn ..
Standalone Worker ( background job, beanstalk,
whenever, resque.... )
Padrino
Pure EventMachine

View Slide

Concurrent is a Myth in Ruby

View Slide

Myths?
Global interpreter lock
Deciphering global interpreter lock
process para$elism
reactor pattern

View Slide

Server architectures & implementation
Architectures Server
Threaded Mongrel, jRuby (torqueBox)
Fork Passenger, Unicorn
Evented Thin, Goliath
Implementation GIL Fiber Thread
MRI 1.8 Y N Green
MRI 1.9 Y Y OS
jRuby N Slow Java Threads
Rubinius 1.x Y N Green, OS
Rubinius 2.0 N Y OS

View Slide

So, how we concurrent in Ruby?

View Slide

EventMachine
EventMachine provides a fast, lightweight %amework for
implementing Ruby programs that can use the network to
communicate with other processes.
It does provide an alternate technique for those applications
requiring better performance, scalability, and discipline over the
behavior of network sockets, than is easily obtainable using the
built-in libraries, especia$y in applications which are
structura$y we$-suited for the event-driven programming
model.

View Slide

Event Machine Components
Deferrable
Make object concurrent
Spawned Process
Erlang style processes
Channel
Pub/Sub
Queue
Queue and Pop

View Slide

It is powerful, but......

View Slide

Callback Hell
r1.callback{
r2.callback{
r3.callback{
r4.callback{
.....
}
}
}
}

View Slide

Hybrid Code
def http_get(url)
f = Fiber.current
http = EventMachine::HttpRequest.new(url).get
# resume fiber once http call is done
http.callback { f.resume(http) }
http.errback { f.resume(http) }
return Fiber.yield
end
EventMachine.run do
Fiber.new{
page = http_get('http://www.google.com/')
puts "Fetched page: #{page.response_header.status}"
if page
page = http_get('http://www.google.com/search?q=eventmachine')
puts "Fetched page 2: #{page.response_header.status}"
end
}.resume
end

View Slide

Hey, Ruby is not functional language..

View Slide

Just grab Node.js , Erlang , Scala , Go , Haske$ ...

View Slide

Reasons
Get advantages %om gems
Easy to integrate with rails, sinatra... etc
Ruby is Beautiful
Ruby is Elegant
Ruby is Agile
Ruby is able to concurrent

View Slide

Good life, don’t you want?

View Slide

Goliath Framework
High performance and concurrent
Based on Event-Machine
Event-Driven
Been Proven ( Pusher.com, PostRank.com )
Easy to use
Low Memory Footprint
2011.03 released

View Slide

Design
Leverage ﬁber(coroutines)
Simple conﬁguration
Middleware support
fu$ asynchronous processing
Readable & Maintainable code

View Slide

Performance
MRI 1.9.2p136 3000 req/s

View Slide

require ‘goliath’

View Slide

Routes with Goliath
class Index < Goliath::API
use Goliath::Rack::Params
get '/', Index
post '/login', Login
get '/overview', Overview
not_found() do run Proc.new{ [200, {'Content-Type' => 'text/plain'}, "Hi. Welcome to API. "] }
end
end

View Slide

Plugins
module Goliath
module Plugin
class Log
def initialize
do_something_here...
end
def run
do_something_here...
end
end
end
end
class YourProject < Goliath::API
use Goliath::Plugin::Log
end

View Slide

Serve Static Files
class YouProject < Goliath::API
use(Rack::Static,
:root => Goliath::Application.app_path("public"),
:urls => ["/favicon.ico", '/stylesheets', '/javascripts', '/images'])
end

View Slide

Session & Cookies
class Auth < Goliath::API
use Rack::Session::Cookie, :secret => 'change me'
use Warden::Manager do |manager|
manager.default_strategies :password
manager.failure_app = Proc.new { |env| [401, {}, 'Access Denied'] }
end
use Authenticator
end

View Slide

Use EM
class Index< Goliath::API
def response
EM.defer{
long_long_execution()
}
end
end
class Channel< Goliath::API
def response
EM.channel.start do
end
end
end
class Timer< Goliath::API
def response
EventMachine::PeriodicTimer(10){
loop_work()
}
end
end
class Queue< Goliath::API
def response
queue = EventMachine::Queue.new
end
end

View Slide

Connect with databases?
Data Mapper
Sequel
ActiveRecord
Swi&
or anything you like....

View Slide

But it is not enough.....

View Slide

Straight forward methods
def response(env)
render :erb
end
end
def response(env)
@res={:status=>"success"}
render :json
end
end

View Slide

Render with erb
def response
@hello="Hello World"
render :erb
end

Everybody.


View Slide

MVC & Generator
$> gm new Balalalala

View Slide

Rake Commands
rake db:create
rake db:rollback
rake db:migrate
rake db:drop
rake db:redo
rake db:console
rake cluster:start
rake cluster:reboot
rake cluster:shutdown

View Slide

Key to Concurrent
Make your each work as sma$ as possible
Don’t block main loop (with em)
Test performance for your each request

View Slide

Some experiences
Use async adapter with your database ex : mysql2
Use better performance gems : sequel, erubis
Database is next bottleneck, have one Nosql for your unimportant data ( feeds, likes )
Use Proc & Block with EM.defer or Thread
Use em-http with your Network request

View Slide

Thanks

View Slide

Enable high coucurrent & availability with Goliath

Enable high coucurrent & availability with Goliath

vonstark

More Decks by vonstark

Other Decks in Technology

Featured

Transcript