Classification

  • This is an educational deployment tutorial: the speaker demonstrates concrete infrastructure and application deployment steps, then explains the operational concepts behind each step. 0:00 22:08

Goal

  • Handle 50 million real HTTP requests per month on a $15/month budget. 0:00
  • The applications are not "hello world": the target system includes about 50 environment variables and about 175 HTTP routes. 0:21
  • The stack includes websockets, realtime chat, server-side rendering, request logging, encryption between clients/proxies/origin, compression, authentication, email, NGINX, Docker, Redis, five SQL databases, uploads, media processing, and storage. 0:39 1:02
  • The target durability for database and media storage is the "99.9999999%" durability level. 1:34

Applications in scope

  • A Java landing-page application with newsletter and contact functionality. 1:52 2:11
  • A Rust backend powering both a web app and an iOS app. 2:11
  • Pagser, a Node.js content-sharing/blogging platform with comments, profile pictures, page photos, attachments, media uploads, and server-side-rendered public pages. 2:30 2:48
  • Weer, a URL shortener that aims to compete on product quality rather than only being a simple redirect service. 3:05 3:20
  • A Next.js marketplace-style app with ads, chat, notifications, large React frontends, code splitting, dark theme, and responsive design. 4:08

Capacity assumptions

  • 50 million requests/month works out to roughly 20 requests/second as a steady average. 5:48
  • The design should account for daily peaks of about 3-5x normal load, so the system should handle roughly 60 requests/second without excessive CPU or RAM pressure. 5:48 6:12
  • The design also budgets for rare 10x spikes, about 200 requests/second. 6:12 6:30
  • The speaker estimates that an application handling this traffic could conservatively correspond to about $300,000/year in revenue, though possibly much more depending on the business. 6:30 6:54

Philosophy

  • The course is framed as a pushback against default cloud-only architectures and cloud egress costs. 1:52 9:30
  • The speaker argues that hybrid cloud plus on-premise systems can reduce cost without necessarily sacrificing availability, performance, or security. 5:31
  • The project is positioned as DevOps education adapted for the AI era, emphasizing realistic systems with real business logic and users. 4:34 8:34