Rails Performance Tips

Richard Huang

@flyerhzm

Where to start

80-90% of the end-user response time is spent on the frontend

High Performance Web Sites by Steve Souders

Frontend Performance Tips

Keith and Mario’s Guide to Fast Websites
From RubyConf Au 2013

1. As less requests as possible

Concatenate CSS & Javascript

Concatenate CSS & JavaScript

Use Css Sprite

Use Css Sprite

2. As small payloads as possible

Minify and gzip

Compress Images

Cookie-less Domains

Host:xinminlabs.com
Referer:https://xinminlabs.com
Cookie:rack.session=BAh7EUkiD3Nlc3Npb25faWQGOgZFVEkiRTJjMjI3NjEzYzQ3
MTZlMmYzZDky%0AMTEzOThkZDdmZjIyYjNiODBlYWI3ZmExYzU1MjRhZjYwN2FhMTBiM
GUwM2YG%0AOwBGSSIJY3NyZgY7AEZJIiUxYzlhMzhlZDBlM2VhNGJhNmFlZDM1MzFjYj
Ri%0AZWYxOQY7AEZJIg10cmFja2luZwY7AEZ7B0kiFEhUVFBfVVNFUl9BR0VOVAY7%0A
AFRJIi1mY2M3ZjA2OGU5N2Q5NzA2YTgwMDdmOTg3MjEwMjJjOThiMjdlYTIw%0ABjsAR
kkiGUhUVFBfQUNDRVBUX0xBTkdVQUdFBjsAVEkiLWRiY2MyMTZiMTBj%0AM2I1MmJkNT
g3ZTBjODIwNTZiMmIyYmE5YjJkNDYGOwBGSSIRYWNjZXNzX3Rv%0Aa2VuBjsAVEkidTA
wRDI4MDAwMDAwZnJFUSFBUU1BUUgxcV9MY2JRbG0yRXh5%0Ad3pqdm1PVjNMZGhNQS5D
SF8xaV94ZDM0X1c5czAwTDgzaHhJV0RaQTBWNW82%0AUFR1clE5cTdUWUtJOTZmUVBoS
2NfUjE3eklsQXp3NHcGOwBUSSIKZW1haWwG%0AOwBUSSIcZmx5ZXJoem0rcHJvZEBnbW
FpbC5jb20GOwBUSSILaWRfdXJsBjsA%0AVEkiSmh0dHBzOi8vbG9naW4uc2FsZXNmb3J
jZS5jb20vaWQvMDBEMjgwMDAw%0AMDBmckVRRUFZLzAwNTI4MDAwMDAxZHlrNkFBQQY7
AFRJIhFpbnN0YW5jZV91%0AcmwGOwBUSSIjaHR0cHM6Ly9saXRhLm15LnNhbGVzZm9yY
2UuY29tBjsAVEki%0ADXVzZXJuYW1lBjsAVEkiHGZseWVyaHptK3Byb2RAZ21haWwuY2
9tBjsAVEki%0ADmlzc3VlZF9hdAY7AFRJIhIxNDU0ODE2NTUxNzE3BjsAVEkiEWRpc3B
sYXlf%0AbmFtZQY7AFRJIhJSaWNoYXJkIEh1YW5nBjsAVEkiEnJlZnJlc2hfdG9rZW4G
%0AOwBUSSJcNUFlcDg2MVRTRVN2V2V1Z193R21UcEcucUlfdDNhbTBpd1JndjFr%0AVD
gyRV8yYjRiQTBMeXFWMzFoVU9Odk1SSEJGMk15V0hzbnZtb2RMVFRYVndr%0AU01UBjs
AVEkiF29yaWdpbmFsX2lzc3VlZF9hdAY7AEZAGw%3D%3D%0A--0e51b9ebf15a845e1f
17b20c0cc947cf8d9b7f60
          

Cookie-less Domains

Host:assets.xinminlabs.com
Referer:https://xinminlabs.com
          

3. As fast resources as possible

Use CDN

4. Other

JavaScript at the Bottom

Backend Performance Tips

How

• Don't guess! Don't guess! Don't guess!
• Find out the bottleneck

Monitor

• New Relic
• Skylight

Monitor

Monitor

Monitor

Reproduce

• Simulate the environment (postgres, memcached, etc.)
• Simulate the data

Reproduce

• New Relic (Developer Mode)
• rack-mini-profiler

Reproduce

Reproduce

Reproduce

Reproduce

Fix

Benchmark

• rails-perftest
• ab / siege

rails-perftest

require 'test_helper'
require 'rails/performance_test_help'

class HomepageTest < ActionDispatch::PerformanceTest
  # Refer to the documentation for all available options
  # self.profile_options = { runs: 5,
  #                          metrics: [:wall_time, :memory],
  #                          output: 'tmp/performance',
  #                          formats: [:flat] }

  test "homepage" do
    get '/'
  end
end
          

rails-perftest

BrowsingTest#test_homepage (58 ms warmup)
        process_time: 63 ms
              memory: 832.13 KB
             objects: 7,882
          

ab / siege

Summarize

• Monitor
• Find out bottleneck
• Reproduce
• Fix
• Benchmark
• Deploy
• Monitor

1. As less requests as possible

N+1 Query

# controller
@comments = Comment.limit(10)

# view
<% @comments.each do |comment| %>
  <%= comment.user.username %> Said: <%= comment.body %>
<% end %>
          

N+1 Query

N+1 Query

Newrelic response time is 12.3ms

N+1 Query

# controller
@comments = Comment.includes(:user).limit(10)

# view
<% @comments.each do |comment| %>
  <%= comment.user.username %> Said: <%= comment.body %>
<% end %>
          

N+1 Query

N+1 Query

Newrelic response time is 6.81ms

N+1 Query

N+1 Query

Counter Cache

# controller
@posts = Post.limit(10)

# view
<% @posts.each do |post| %>
  <%= post.title %> has <%= post.comment.size %> comments.
<% end %>
          

Counter Cache

Counter Cache

Newrelic response time is 12.3ms

Counter Cache

# migration
add_column :posts, :comments_count, :integer
Post.all.each do |post|
  Post.reset_counters(post.id, :comments)
end

# model
class Comment < ActiveRecord::Base
  belongs_to :post, counter_cache: true
end
          

Counter Cache

Counter Cache

Newrelic response time is 7.04ms

Counter Cache

Counter Cache

Bonus

Use Group

# model
class Comment
  belongs_to :post
  scope :approved, -> { where(approved: true) }
end

class Post
  has_many :comments
  def average_rating
    comments.approved.average(:rating)
  end
end
          

Use Group

# controller
@posts = Post.limit(10)

# view
<% @posts.each do |post| %>
  <%= post.title %> average rating is
  <%= post.average_rating %> stars
<% end %>
          

Use Group

Use Group

Newrelic response time is 17.4ms

Use Group

# Gemfile
gem 'eager_group'

# model
class Post < ActiveRecord::Base
  has_many :comments
  define_eager_group :average_rating, :comments, :average, :rating,
-> { approved }
end

# controller
@posts = Post.eager_group(:average_rating).limit(10)
          

Use Group

Use Group

Newrelic response time is 7.48ms

Use Group

Use Group

Bonus

Multi Inserts

# controller
10.times do
  Post.create title: Faker::Lorem.sentence,
              body: Faker::Hipster.paragraph,
              user_id: rand(100) + 1
end
          

Multi Inserts

Multi Inserts

Newrelic response time is 35.4ms

Multi Inserts

# Gemfile
gem 'activerecord-import'

# controller
posts = []
10.times do
  posts << Post.new(title: Faker::Lorem.sentence,
                    body: Faker::Hipster.paragraph,
                    user_id: rand(100) + 1)
end
Post.import posts
          

Multi Inserts

Multi Inserts

Newrelic response time is 11.6ms

Multi Inserts

Multi Inserts

Multi Updates

Post.where("created_at < ?", 10.years.ago).update_all(archive: true)
          

Multi Deletes

Post.where("created_at < ?", 10.years.ago).destroy_all

Post.where("created_at < ?", 10.years.ago).delete_all
          

2. As small payloads as possible

Select Data You Really Need

# controller
@posts = Post.limit(10)

# view
<% @posts.each do |post| %>
  <%= post.title %>
<% end %>
          

Select Data You Really Need

Select Data You Really Need

Newrelic response time is 4.86ms

Select Data You Really Need

# controller
@posts = Post.select('title').limit(10)
          

Select Data You Really Need

Select Data You Really Need

Newrelic response time is 4.54ms

Select Data You Really Need

3. As fast resources as possible

Use Cache

# controller
@posts = Post.limit(10)

# view
<% @posts.each do |post| %>
  <%= post.user.username %> said <%= post.title %>
<% end %>
          

Use Cache

Use Cache

Newrelic response time is 12.2ms

Use Cache

# Gemfile
gem 'dalli'

# config/application.rb
config.cache_store = :mem_cache_store

# view
<% @posts.each do |post| %>
  <% cache post do %>
    <%= post.user.username %> said <%= post.title %>
  <% end %>
<% end %>
          

Use Cache

Use Cache

Newrelic response time is 9.97ms

Use Cache

Bonus

Rails.cache.read_multi

Use ElasticSearch for search

Use Redis

Use SSD

4. Other

Use Index

# controller
@posts = Post.order('created_at desc').limit(10)

# view
<% @posts.each do |post| %>
  <%= post.title %>
<% end %>
          

Use Index

SELECT  "posts".* FROM "posts"  ORDER BY created_at desc LIMIT 10
          

Use Index

Newrelic response time is 21.5ms

Use Index

# migration
def change
  add_index :posts, :created_at
end
          

Use Index

Newrelic response time is 4.95ms

Use Index

Use Index

Optimize JSON Rendering

# controller
@posts = Post.limit(10)
render json: @posts
          

Optimize JSON Rendering

Newrelic response time is 6.28ms

Optimize JSON Rendering

# Gemfile
gem 'oj'
gem 'oj_mimic_json'
          

Optimize JSON Rendering

Newrelic response time is 5.76ms

Optimize JSON Rendering

Optimize JSON Rendering

Memory Optimize

shave off 1,114 string objects on every request

shave off 34,299 objects on every request

Find in batch

Person.where("age > 21").each do |person|
  person.party_all_night!
end
          

Person.where("age > 21").find_each do |person|
  person.party_all_night!
end
          

Find in batch

• Send multiple sql requests
• But reduce memory usage

No tip is always true

Review

• Frontend performance tuning does first
• Backend performance tuning is important

Review

• As less requests as possible
• As small payloads as possible
• As fast resources as possible
• Other

Review

• Monitor
• Find out bottleneck
• Reproduce
• Fix
• Benchmark
• Deploy
• Monitor

Demos Code

Thank You

Richard Huang

@flyerhzm