The Trail to Scale Without Fail: Rails?

on YouTube ↗01h 03m

" summary

Cloudinary runs 15 billion requests per day and relies on a CDN to serve ~95% of traffic, a tiny Go service called seget (Go) handling 85% of cache misses, and Rails for the IO/CPU layers that see about 150M and 125M requests/day respectively. Scaling uses per-cloud sharding, a near‑real‑time cache for multi‑region, a Scala Lobster lock to avoid duplicate transforms, and a Go‑based fair queuing system; these pieces reduce cost and, crucially, actually work in production.

AI-written summary · about

" chapters (25)

1. 0:00Start
2. 1:53Requests per Second
3. 5:05What Does cloudinary Actually Do
4. 9:12Image Formats
5. 10:01Cdns
6. 12:04The Delivery Pipeline
7. 14:06Auto Scaling
8. 15:35Request Life Cycle
9. 17:05I O Layer
10. 19:20Cdn
11. 21:50Invalidation System
12. 24:30Advantages of Layers
13. 25:32Security
14. 26:23Sharding
15. 27:08Vertical Partitioning
16. 27:37Vertically Partitioning
17. 35:06Regions at Cloudinary
18. 42:18Best Effort Locking System
19. 45:12Downtime
20. 45:34How Do You Scale Your Locking System
21. 49:19Rate Limits
22. 49:52Fair Queuing System
23. 56:12Eager Transformations
24. 56:41Human Factors
25. 58:03Cloudinary on Rails

" related talks

• 27mRails Orbiting the cloud: Mastering the scalability with KubernetesRahul Mahale · RubyConf India 2023
• 38mRails in the CloudIan Warshak · Lone Star Ruby Conference 2009
• 28mSimplifying at Scale: 7 Years of Rails ArchitectureAlex Coomans · RailsConf 2025
• 44mIntroduction to application scalingLiam Wiltshire · PHP UK Conference 2017
• 31mA RAILS PERFORMANCE GUIDEBOOKCristian Planas · Euruko 2023
• 40mHow Shopify Scales RailsJohn Duff · RailsConf 2013
• 37mScaling Rails for Black Friday and Cyber MondayChristian Joudrey · RailsConf 2015
• 31mA Rails performance guidebook: from 0 to 1B requests/dayCristian Planas · RubyConf TH 2022