Cache = Cash! 2.0

RailsConf 2025 talk

Transcript

1. Cache = Cash! 2.0 Stefan Wintermeyer @wintermeyer
2. Feel free to ask questions anytime!
5. 2013 2025 v1.0 v2.0
6. Code Examples
7. Code Examples Agent Prompts
8. Kleinvieh macht auch Mist. As we Germans put it: "Even small animals make a mess." 
 Small steps can lead to something big.
9. Stuff that I didn’t think about in 2013
10. CPU cache
12. Cache vs. CPU Ruby: result = 1 + 1 Ruby: initials = "#{'Stefan'[0]} #{'Wintermeyer'[0]}"
13. Program 1: Add 1 + 1 ──────────────────── mov al, 1 ; 1 cycle add al, 1 ; 1 cycle mov [result], al ; 3 cycles ; = 5 cycles total
14. Program 2: Make "S.W." Initials ──────────────────────────────── mov al, [Stefan] ; 3 cycles mov [init], al ; 3 cycles mov al, [Wintermeyer]; 3 cycles mov [init+2], al ; 3 cycles ; = 12 cycles total
15. Caching: One size doesn’t fi t all! The very same code can perform totally di ff erent on an other hardware!!!
16. 2013 setup v1.0
17. Raspberry Pi (35 USD)
18. Cart state string LineItem price decimal (8,2) quantity integer Category name string Product description text name string price decimal (8,2) DiscountGroup discount integer name string User email string ∗ encrypted_password string ∗ first_name string last_name string Rating value integer
19. Bob‘s cart
20. 0 30 60 90 120 Vanilla Fragment Caching HTTP Cache Preheater Page Caching
21. 2025 setup v2.0
28. Messure Improvements
30. $ bin/rails benchmark:shop Users: 300 (240 browsers, 60 buyers) Result: Total Time: 30289 ms ├─ Views: 20873 ms ├─ DB: 9416 ms Baseline
31. $ bin/rails benchmark:shop Users: 300 (240 browsers, 60 buyers) Result: Total Time: 30289 ms ├─ Views: 20873 ms ├─ DB: 9416 ms Baseline
32. select
34. stefan@SW-MacBook-Air-M4 shop3 % git diff --unified=0 | grep -E '^\+\+\+|^@@|^\+[^+]|^-[^-]' +++ b/app/controllers/application_controller.rb @@ -26 +26 @@ class ApplicationController < ActionController::Base - @current_cart = Cart.active.find_by(id: session[:cart_id]) + @current_cart = Cart.active.select(:id, :user_id, :state, :session_id).find_by(id: session[:cart_id]) @@ -34 +34 @@ class ApplicationController < ActionController::Base - cart = Cart.active.find_by(user: current_user) + cart = Cart.active.select(:id, :user_id, :state, :session_id).find_by(user: current_user) @@ -44 +44 @@ class ApplicationController < ActionController::Base - @session_cart = Cart.active.find_by(id: session[:cart_id]) + @session_cart = Cart.active.select(:id, :user_id, :state, :session_id).find_by(id: session[:cart_id]) @@ -66 +66 @@ class ApplicationController < ActionController::Base - session_cart = Cart.find_by(id: session[:cart_id]) + session_cart = Cart.select(:id, :user_id, :state).find_by(id: session[:cart_id]) +++ b/app/controllers/carts_controller.rb @@ -4 +4 @@ class CartsController < ApplicationController - @line_items = @cart ? @cart.line_items.includes(:product) : [] + @line_items = @cart ? @cart.line_items.preload(product: proc { select(:id, :name, :tagline, :price) }) : [] +++ b/app/controllers/home_controller.rb @@ -6 +6 @@ class HomeController < ApplicationController - .select("products.*, AVG(ratings.value) as avg_rating, COUNT(ratings.id) as ratings_count") + .select("products.id, products.name, products.tagline, products.price, products.weight, products.expiration_date, products.category_id, AVG(ratings.value) as avg_rating COUNT(ratings.id) as ratings_count") @@ -11 +11 @@ class HomeController < ApplicationController - .includes(:category) + .preload(category: proc { select(:id, :name) }) +++ b/app/controllers/line_items_controller.rb @@ -5 +5 @@ class LineItemsController < ApplicationController - product = Product.find(params[:product_id]) + product = Product.select(:id, :price).find(params[:product_id]) +++ b/app/controllers/products_controller.rb @@ -3 +3,4 @@ class ProductsController < ApplicationController - @product = Product.includes(:category, :ratings).find(params[:id]) + @product = Product + .select(:id, :name, :tagline, :price, :weight, :dimensions, :expiration_date, :details, :category_id, :created_at, :updated_at) + .preload(category: proc { select(:id, :name) }) + .find(params[:id]) +++ b/app/models/cart.rb @@ -22 +22 @@ class Cart < ApplicationRecord - current_item = line_items.find_by(product: product) + current_item = line_items.select(:id, :product_id, :quantity, :price).find_by(product: product) @@ -41 +41 @@ class Cart < ApplicationRecord - line_items.find_by(product: product)&.quantity || 0 + line_items.select(:quantity).find_by(product: product)&.quantity || 0 +++ b/app/views/shared/_product_cart_controls.html.erb @@ -3 +3 @@ - <% line_item = current_cart.line_items.find_by(product: product) %> + <% line_item = current_cart.line_items.select(:id, :quantity).find_by(product: product) %>
35. @@ -11 +11 @@ class HomeController < ApplicationController - .includes(:category) + .preload(category: proc { select(:id, :name) })
36. Optimize Datatypes
39. select + optimized datatypes = - 3% 0 10.000 20.000 30.000 40.000 Baseline select + datatype 0 29.380 30.289
40. Indexes
43. Indexes = - 6% 0 10.000 20.000 30.000 40.000 Baseline select + datatype Index 0 27.617 29.380 30.289
44. class AddNameIndexesToUsers < ActiveRecord::Migration[8.0] def change add_index :users, :last_name, name: "ln" add_index :users, [:last_name, :first_name], name: "ln_and_fn" end end Index Pro Tipro-Tip
45. Fragment Caching and Russian Doll
47. Fragment Caching
48. Russian Doll
50. Fragment Caching = - 60% 0 10.000 20.000 30.000 40.000 Baseline select + datatype Index Fragment Caching 0 11.046 27.617 29.380 30.289
51. HTTP Caching
52. HTTP caching still uses the controller but doesn’t always render the HTML Less computing on the server and less/faster payload to the web client.
53. etag is a fi ngerprint
54. > curl -I http://0.0.0.0:3000/ HTTP/1.1 200 OK Content-Type: text/html; charset=utf-8 X-Ua-Compatible: IE=Edge Etag: "9a779b80e4b0ac3c60d29807e302deb7" [...] > curl -I http://0.0.0.0:3000/ HTTP/1.1 200 OK Content-Type: text/html; charset=utf-8 X-Ua-Compatible: IE=Edge Etag: "fa8fc1e981833a6885b583d351c4d823"
55. > curl -I http://0.0.0.0:3000/ HTTP/1.1 200 OK Content-Type: text/html; charset=utf-8 X-Ua-Compatible: IE=Edge Etag: "9a779b80e4b0ac3c60d29807e302deb7" [...] > curl -I http://0.0.0.0:3000/ HTTP/1.1 200 OK Content-Type: text/html; charset=utf-8 X-Ua-Compatible: IE=Edge Etag: "fa8fc1e981833a6885b583d351c4d823"
58. > curl -I http://0.0.0.0:3000/ -c cookies.txt HTTP/1.1 200 OK Etag: "4d348810e69400799e2ab684c0ef4777" > curl -I http://0.0.0.0:3000/ -b cookies.txt HTTP/1.1 200 OK Etag: "4d348810e69400799e2ab684c0ef4777" The cookie is needed for the CSRF-Token.
59. > curl -I http://0.0.0.0:3000/ -b cookies.txt --header 'If-None-Match: "4d348810e69400799e2ab684c0ef4777"' HTTP/1.1 304 Not Modified Etag: "4d348810e69400799e2ab684c0ef4777" 304!
60. last_modi fi ed Same idea like etag but with a timestamp.
63. Conditional Request - Using If-Modified-Since # Use the Last-Modified value from the first request curl -I http://localhost:3000/ \ -H "If-Modified-Since: Mon, 07 Jul 2025 10:30:45 GMT" # If content hasn't changed, you'll get: HTTP/1.1 304 Not Modified Last-Modified: Mon, 07 Jul 2025 10:30:45 GMT
64. HTTP Caching = - 23% 0 10.000 20.000 30.000 40.000 Baseline select + datatype Index Fragment Caching HTTP Caching 0 8.505 11.046 27.617 29.380 30.289
65. 3 Autobahn!
66. The fastest page is delivered by Nginx without ever asking Rails.
67. ᵓᴷᴷ Gem fi le ᵓᴷᴷ [...] ᵓᴷᴷ public │ ᵓᴷᴷ 404.html │ ᵓᴷᴷ 422.html │ ᵓᴷᴷ 500.html │ ᵓᴷᴷ favicon.ico │ └── robots.txt ᵓᴷᴷ [...] That‘s already done for the fi les in the public directory.
68. Static copies of products#show 😱
71. stefan@Mac shop3 % tree public public ├── 400.html ├── 404.html ├── 406-unsupported-browser.html ├── 422.html ├── 500.html ├── icon.png ├── icon.svg ├── products │ ├── 501.html │ ├── 501.html.br │ ├── 502.html │ ├── 502.html.br │ ├── 503.html Disk space is cheap!
72. How does nginx know which one to use?!?! 😱 😱 😱
77. Total = - 89% 0 10.000 20.000 30.000 40.000 Baseline select + datatype Index Fragment Caching HTTP Caching Autobahn 3.572 8.505 11.046 27.617 29.380 30.289
78. Why is speed so important?
79. Usability Engineering 101 Delay User reaction 0 - 100 ms Instant 100 - 300 ms Feels sluggish 300 - 1000 ms Machine is working... 1 s+ Mental context switch 10 s+ I'll come back later... Stay under 250 ms to feel "fast". Stay under 1000 ms to keep users attention. @igrigorik
80. Yo ho ho and a few billion pages of RUM How speed affects bounce rate @igrigorik
81. Let’s compare two webpages
83. https://railsconf.org/schedule/
84. https://www.webpagetest.org/result/250708_ZiDc5J_614/ https://railsconf.org/schedule/
85. https://www.webpagetest.org/result/250708_YiDcXQ_664/ https://wintermeyer-consulting.de
86. railsconf.org/schedule wintermeyer-consulting.de Time to First Byte .316s .825s Start Render 1.400s 0.900s Speed Index 3.097s 0.902s
87. wintermeyer-consulting.de Stefan Wintermeyer @wintermeyer