Siyao Zhao, Haoyu Gu, and Ali José Mashtizadeh, University of Waterloo
This paper presents Schedulable Kqueue (SKQ), a new design to FreeBSD Kqueue that improves application tail latency and low-latency throughput. SKQ introduces a new scalable architecture and event scheduling. We provide multiple scheduling policies that improve cache locality and reduce workload imbalance. SKQ also enables applications to prioritize processing latency-sensitive requests over regular requests.
In the RocksDB benchmark, SKQ reduces tail latency by up to 1022× and extends the low-latency throughput by 27.4×. SKQ also closes the gap between traditional OS kernel networking and a state-of-the-art kernel-bypass networking system by 83.7% for an imbalanced workload.
Open Access Media
USENIX is committed to Open Access to the research presented at our events. Papers and proceedings are freely available to everyone once the event begins. Any video, audio, and/or slides that are posted after the event are also free and open to everyone. Support USENIX and our commitment to Open Access.
author = {Siyao Zhao and Haoyu Gu and Ali Jos{\'e} Mashtizadeh},
title = {{SKQ}: Event Scheduling for Optimizing Tail Latency in a Traditional {OS} Kernel},
booktitle = {2021 USENIX Annual Technical Conference (USENIX ATC 21)},
year = {2021},
isbn = {978-1-939133-23-6},
pages = {759--772},
url = {https://www.usenix.org/conference/atc21/presentation/zhao-siyao},
publisher = {USENIX Association},
month = jul
}