TY - GEN
T1 - Learning relaxed belady for content distribution network caching
AU - Song, Zhenyu
AU - Berger, Daniel S.
AU - Li, Kai
AU - Lloyd, Wyatt
N1 - Funding Information:
This research is partly funded by a Comcast Innovation grant, a Princeton University fellowship, and a Facebook research grant. We are grateful to our anonymous reviewers, our shepherd Rebecca Isaacs, Amit Levy, and David Liu whose extensive comments substantially improved this work. We also thank Fei Gao, Qizhe Cai, Rong Huang, Jeffrey Helt, and Jen-nifer Lam who contributed to the project at different stages.
Publisher Copyright:
© Proc. of the 17th USENIX Symposium on Networked Systems Design and Impl., NSDI 2020. All rights reserved.
PY - 2020
Y1 - 2020
N2 - This paper presents a new approach for caching in CDNs that uses machine learning to approximate the Belady MIN (oracle) algorithm. To accomplish this complex task, we propose a CDN cache design called Learning Relaxed Belady (LRB) to mimic a Relaxed Belady algorithm, using the concept of Belady boundary. We also propose a metric called good decision ratio to help us make better design decisions. In addition, the paper addresses several challenges to build an end-to-end machine learning caching prototype, including how to gather training data, limit memory overhead, and have lightweight training and prediction. We have implemented an LRB simulator and a prototype within Apache Traffic Server. Our simulation results with 6 production CDN traces show that LRB reduces WAN traffic compared to a typical production CDN cache design by 4-25%, and consistently outperform other state-of-the-art methods. Our evaluation of the LRB prototype shows its overhead is modest and it can be deployed on today's CDN servers.
AB - This paper presents a new approach for caching in CDNs that uses machine learning to approximate the Belady MIN (oracle) algorithm. To accomplish this complex task, we propose a CDN cache design called Learning Relaxed Belady (LRB) to mimic a Relaxed Belady algorithm, using the concept of Belady boundary. We also propose a metric called good decision ratio to help us make better design decisions. In addition, the paper addresses several challenges to build an end-to-end machine learning caching prototype, including how to gather training data, limit memory overhead, and have lightweight training and prediction. We have implemented an LRB simulator and a prototype within Apache Traffic Server. Our simulation results with 6 production CDN traces show that LRB reduces WAN traffic compared to a typical production CDN cache design by 4-25%, and consistently outperform other state-of-the-art methods. Our evaluation of the LRB prototype shows its overhead is modest and it can be deployed on today's CDN servers.
UR - http://www.scopus.com/inward/record.url?scp=85091835547&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85091835547&partnerID=8YFLogxK
M3 - Conference contribution
AN - SCOPUS:85091835547
T3 - Proceedings of the 17th USENIX Symposium on Networked Systems Design and Implementation, NSDI 2020
SP - 529
EP - 544
BT - Proceedings of the 17th USENIX Symposium on Networked Systems Design and Implementation, NSDI 2020
PB - USENIX Association
T2 - 17th USENIX Symposium on Networked Systems Design and Implementation, NSDI 2020
Y2 - 25 February 2020 through 27 February 2020
ER -