SMUFF: Towards Line Rate Wi-Fi Direct Transport with Orchestrated On-device Buffer...
Talkat 21st USENIX Symposium on Networked Systems Design and Implementation (NSDI '24) · 2024-04-16..04-18·
summary
SMUFF implements orchestrated on-device buffer management that monitors every buffer on the sender–receiver path and dynamically backlogs just enough data in the bottleneck to avoid underflow, removing conventional congestion control and using reduced ACK frequency, negative/ selective ACKs and packet aggregation. Deployed on four phone models, it reaches up to 94.7% practical line rate, averages ~89% on Wi‑Fi5 and 91% on Wi‑Fi6, yields 22.6% higher throughput, 37% lower CPU and 15% lower power versus prior solutions.
AI-written summary · about
chapters (9)
- 0:00SMUFF and Wi-Fi Direct
- 2:47Peer-to-peer Wi-Fi Direct
- 5:40Wi-Fi Direct throughput gaps
- 7:28SMUFF design goals
- 9:17On-device buffers for line rate
- 13:34Sending rates with buffer feedback
- 14:56SMUFF implementation
- 15:17Performance evaluation
- 15:59Takeaways and future work
related talks
45mReducing Network Latency: Innovations for a Faster Internet
43mWi-Fi 6 (Formerly IEEE 802.11ax) Deep Dive
17mEnabling Users to Control their Internet
19mDChannel: Accelerating Mobile Applications With Parallel High-bandwidth...
17mBLADE: Adaptive Wi-Fi Contention Control for Next-Generation Real-Time Communication
44mLinux Networking with MANA: RX Path Optimization and Netsh...
16mPacket Order Matters! Improving Application Performance by Deliberately Delaying Packets
23mBuilding a network stack for optimal throughput / low-latency trade-offs