Concurrent Web Server | https://speakerdeck.com/exaspark/how-to-make-our-concurrency-tools-suck-less-at-servers-dot-com

README.md

Concurrent Server

Requirements

• puma
• sinatra
• typhoeus

Installation

bundle install

Running server

PORT=3003 REQUESTS_FROM_SERVER=5 make run_server

Running experiment

time PORT=3003 REQUESTS_TO_SERVER=4 make run_experiment

Results

Workers	Threads	Concurrent requests from server	Concurrent requests to server	Time
2	2	1	1	1.6
2	2	1	4	1.6
2	2	1	8	4.5
2	2	1	16	6.4
2	2	1	32	13.5
2	2	5	1	1.6
2	2	5	4	1.6
2	2	5	8	4.5
2	2	5	16	6.9
2	2	5	32	14.1

Conclusion

• Number of concurrent requests from server doesn't affect performance.
• Time of concurrent requests to server which fit thread pool size is constant.
• Time of concurrent requests to server which doesn't fit thread pool size grows linearly.

app.rb

require 'sinatra'
require 'typhoeus'
require 'typhoeus/adapters/faraday'
require 'json'

class App < Sinatra::Base
  REQUESTS_FROM_SERVER = ENV['REQUESTS_FROM_SERVER'].to_i

  get '/' do
    responses = make_concurrent_requests
    sleep 1
    responses.map(&:status).to_json
  end

  private

  def make_concurrent_requests
    responses = []
    conn = Faraday.new(url: "https://google.com") { |faraday| faraday.adapter(:typhoeus) }

    conn.in_parallel do
      REQUESTS_FROM_SERVER.times { responses << conn.get('/') }
    end

    responses
  end
end

chart.png

config.ru

require_relative './app'

configure do
  set :server, :puma
end

puts "Concurrent request count from server: #{App::REQUESTS_FROM_SERVER}"

run App

Gemfile

source 'https://rubygems.org'

gem 'puma'

gem 'sinatra'

gem 'faraday'
gem 'typhoeus'

group :development do
  gem 'pry'
end

Makefile

run_server:
	bundle exec puma -p $(PORT) -t 0:2 -w 2 --preload config.ru
run_experiment:
	for n in {1..$(REQUESTS_TO_SERVER)}; do echo 'fetching..'; curl -s -o /dev/null http://0.0.0.0:$(PORT) & done; wait