DMon: Efficient Detection and Correction of Data Locality Problems Using Selective

on YouTube ↗14m

" summary

Selective profiling collects runtime data selectively and incrementally to locate and repair data-locality bugs with low overhead; the prototype DMon automatically identifies bad access patterns, applies targeted code or data-layout fixes, and runs in production with 1.36% overhead on average. Evaluations show DMon’s optimizations yield 16.83% average speedup (up to 53.14%), and speed up PostgreSQL by 6.64% average (up to 17.48%), demonstrating practical, deployable improvements.

AI-written summary · about

" chapters (5)

1. 0:00Introduction to DMon
2. 2:17Memory locality problem focus
3. 5:16Demo on memory access latency
4. 10:10Real-world applications and results
5. 12:24State of the problem and conclusions