Bridging Storage and Execution: A Semantic Virtual Bus for On-Demand Application Streaming
summary
STREAMBUS implements a semantic virtual storage bus that turns remote storage into an execution-aware layer by combining a dual-mode transmission (fast on-demand fetch plus slow background prefetch) with a thread-aware Markov-chain model for fine-grained block prediction. It is implemented beneath the filesystem and integrated into Android, skips full downloads and installation, and delivers near-native performance: 15–40% better per-page latency than NVMe in common desktop cases, startup overheads typically below 40% on Wi‑Fi mobile, and 45–78% faster end-to-end on desktop versus native install.
AI-written summary · about
chapters (13)
- 0:00Stream Bus concept
- 1:44Application delivery problems
- 3:31Remote execution limitations
- 4:54On-demand streaming challenges
- 5:59Stream Bus design
- 7:23Execution-aware prediction model
- 9:08Code block prediction model
- 10:15Android integration setup
- 11:19Cache hit improvements
- 12:22Desktop vs mobile performance
- 13:29Startup time comparisons
- 14:15Network robustness
- 15:19Conclusion
related talks
43mSmarter data storage on Android
40mAndroid high performance baby
08h 39mLivestream | Day 2, theater 2
25mReducing Android Boot Times: Evaluating Proactive File Prefetching
24mLightweight Realtime & Offline Data Storage on Android without a Database
14mHelp your players spend less time waiting and more time playing with Play as you download
39mEffective Data Prefetching on Background
31mSoftware Defined Vehicle and Android Automotive OS