Conference proceeding
KeepON: Supporting Deterministic Traffic on Standard NICs
23RD USENIX SYMPOSIUM ON NETWORKED SYSTEMS DESIGN AND IMPLEMENTATION, NSDI 2026, pp.209-236
01/01/2026
Abstract
Networked mission-critical applications (e.g., avionics control and industrial automation) demand deterministic packet transmissions to meet stringent sensing and control timing requirements. While specialized infrastructures such as Time-Triggered Ethernet and Time-Sensitive Networking (TSN) ensure deterministic data delivery across switches, end devices still require specialized NICs (e.g., TSN NICs or NVIDIA Mellanox) to eliminate endpoint indeterminism. However, deploying such NICs at every endpoint is costly and hinders compatibility with legacy systems. To address this challenge, we propose KeepON, a novel software-based driver model that enables deterministic packet transmission on commodity NICs. The core idea is to continuously transmit fixed-size placeholder packets, establishing a predictable transmission pattern. Mission-critical packets are then precisely inserted into this stream by replacing placeholders at their scheduled transmission slots, ensuring timing accuracy. The placeholder packets are efficiently dropped at the first-hop switch, avoiding negative impacts on network performance. We prototype KeepON by modifying the standard NIC driver of a Raspberry Pi(1), and integrate it into a real-world TSN testbed. Experimental results show that KeepON achieves up to 130x improvement in scheduling accuracy compared to the default driver, and 2.1x improvement over a hardware-based solution.
Details
- Title: Subtitle
- KeepON: Supporting Deterministic Traffic on Standard NICs
- Creators
- Chuanyu Xue - University of ConnecticutTianyu Zhang - University of IowaAndrew Loveless - Johnson Space CenterSong Han - University of Connecticut
- Resource Type
- Conference proceeding
- Publication Details
- 23RD USENIX SYMPOSIUM ON NETWORKED SYSTEMS DESIGN AND IMPLEMENTATION, NSDI 2026, pp.209-236
- Publisher
- Usenix Assoc
- Number of pages
- 28
- Grant note
- CNS-2008463; CNS-2432533 / National Science Foundation; National Science Foundation (NSF) 80NSSC19K1076 / NASA STRI Resilient Extraterrestrial Habitats Institute (RETHi)
- Language
- English
- Date published
- 01/01/2026
- Academic Unit
- Computer Science
- Record Identifier
- 9985219316902771
Metrics
1 Record Views