Differentiating Cache Files for Fine-grain Management to Improve Mobile Performance
Talkat 12th USENIX Workshop on Hot Topics in Storage and File Systems (HotStorage '20) · 2020-07-13..07-14
summary
Analysis of two hours of traces across 10 apps (one case: Facebook generates 1.2 GB of cache) shows cache files vary by lifetime, size and read pattern: writes are ~100× larger than reads and >50% of cache bytes come from long-lived files. A hybrid memory+flash framework FCFM classifies files into burn-after-reading, transient, and long-living, keeps transient ones in RAM and discards burn-after-reading. Evaluation on YouTube cuts read-back and write I/O substantially, with ~2% extra re-downloads as the measured penalty.
AI-written summary · about
chapters (10)
- 0:00Intro
- 0:29Executive summary
- 3:39Cache File Analysis Experimental setup
- 4:05Cache File Analysis at Application-level • Read and write amount of cache files in flash storage
- 4:49Cache File Analysis at File-level
- 6:46Cache File Analysis at Read Pattern-level Read pattern of the ten most read cache files for each application.
- 8:02Fine-grain cache file management (FCFM)
- 9:23Preliminary evaluation and analysis Preliminary evaluation results video cache of YouTube
- 11:01Discussion Topics
- 11:58Conclusion