Caching Entire Systems without Invalidation

on YouTube ↗34m

" summary

Argues systems become cacheable by resolving all variable inputs up front and splitting into a state gathering layer and an immutable zone so downstream functions always see fixed values. Shows a concrete days‑old example, then two patterns for big datasets: the timestamped data pattern (index with timestamps + immutable document store) and a snapshot service; Bloomberg uses the append‑only document pattern for a mortgage‑backed analytics pipeline and reports it works, with cache keys including service versions.

AI-written summary · about

" chapters (15)

1. 0:00Introduction to architecture
2. 1:41The role of SRE
3. 3:31Defining statelessness
4. 5:57Caching benefits
5. 7:11Hard problems in caching
6. 10:33Interface design for caching
7. 13:45Resilient cache behavior
8. 15:36Key takeaways
9. 17:05Handling large databases
10. 18:59Timestamp data pattern
11. 21:15Object store pattern
12. 22:19Snapshot data pattern
13. 25:23Benefits of timestamp data
14. 27:11Managing software versions
15. 28:52Conclusion and Q&A

" related talks

• 31mApplication Cache: DouchebagJake Archibald · Mobilism 2012
• 28mNever Wait for an API Again: Unlocking CacheStorage for Blazing SpeedNaveh Sadeh · ReactNext 2025
• 31mThe Opposite of Spaghetti Code: Building for UnderstandingBrandon W Maister · RustConf 2018
• 34mImmutability: Putting The Dream Machine To WorkDavid Nolen · GOTO Copenhagen 2014
• 49mReactive microsystemsIgnasi Marimon · Software Crafters Barcelona 2018
• 01h 06mPatterns of Basecamp's Application ArchitectureDavid Heinemeier Hansson · RailsConf 2013
• 28mPersistent data structures - good for UIDanie Roux · RubyFuza 2014
• 32mThere and Back Again: Grokking State and DataMichel Weststrate · React Amsterdam 2018