High Performance Architecture
summary
Demonstrates that call-return APIs cause O(depth × width) temporary allocations (example: object description recursion that quickly exhausts memory and forces swapping), and shows two fixes: replace return-value composition with pipe and filter style writers (write into a provided buffer to eliminate copies), and redesign a GTFS importer by bit-packing times into 32-bit records, using bucket sort and mmap for zero-copy reads. Result: 22 minutes → <1 second import, 0.5 s queries → 58 µs (≈8,000×), 130 MB CSV → 12 MB binary.
AI-written summary · about
chapters (5)
- 0:00Introduction and background
- 1:29Performance and architecture overview
- 6:10Call return architectural style
- 14:29Layering and hexagonal architecture
- 25:11Key takeaways and conclusion
related talks
34mFast Code Without Breaking Things
01h 03mTurbocharged: Writing High-Performance C# and .NET Code
48mOptimising a small real-world C++ application
37mCausal's 1000x Spreadsheet; Performance beyond profiling
49mSee the tries for the trees
23mFast by default: algorithmic performance optimization in practice
40mLudicrous Speed: Designing For Performance on the JVM
33mReaching Ludicrous Speed