SRE Classroom - How to Design a Distributed System in 3 Hours
summary
Workshop converts a sketch into a runnable design: given a million users uploading 50 photos/day of 4 MB each and thumbnails of 256 KB, the exercise uses the provided storage spec and SLOs 99.9% SLO 250ms thumbnails 200ms full images to derive traffic and resource numbers like 579 QPS uploads and ~58 QPS for searches/downloads. Design splits into upload, thumbnail, download and index backed by globally replicated storage, sizes ~13 servers per DC and ~54 servers global, applies n+2 redundancy and caching, and validates the example solution meets SLOs while noting a 1 Gb vs 10 Gb NIC miscalculation.
AI-written summary · about
chapters (18)
- 0:00Intro
- 2:37Requirements
- 7:23Scaling in Distributed Systems
- 9:27Failure Domains
- 11:18Dealing with Failure
- 12:34Defending Against Failure
- 15:14Achieving Reliability
- 19:06Consistency - CAP
- 25:07Problem Statement
- 29:40Available Hardware
- 38:48Example Solution
- 40:30Downloading a Full Size Picture
- 46:15Upload Service - Bandwidth
- 49:31Thumbnail Service - Timing
- 51:01Download Service - Bandwidth
- 55:56Load Balancer - Bandwidth
- 56:44Footprint - Summary
- 59:06Global Footprint
related talks
30mDistributed system design
13mExplaining Distributed Systems Like I'm 5
34mDistributed Systems - What Can Go Wrong and How to Get It Right
38mThe SRE’s Crystal Ball: Predicting System Performance with...
31mThe Design of a Modern Distributed Object Storage
53mCheaper, More Reliable, Simpler
34mDistributed systems 102
29mHorrors of Distributed Systems