Driving cache replacement with ML-based LeCaR

Giuseppe Vietri; Liana V. Rodriguez; Wendy A. Martinez; Steven Lyons; Jason Liu; Raju Rangaswami; Ming Zhao; Giri Narasimhan

Driving cache replacement with ML-based LeCaR

Giuseppe Vietri, Liana V. Rodriguez, Wendy A. Martinez, Steven Lyons, Jason Liu, Raju Rangaswami, Ming Zhao, Giri Narasimhan

Engineering, Ira A. Fulton Schools of (IAFSE)

Research output: Contribution to conference › Paper › peer-review

62 Scopus citations

Abstract

Can machine learning (ML) be used to improve on existing cache replacement strategies? We propose a general framework called LeCaR that uses the ML technique of regret minimization to answer the question in the affirmative. We show that the LeCaR framework outperforms ARC using only two fundamental eviction policies, LRU and LFU, by more than 18x when the cache size is small relative to the size of the working set.

Original language	English (US)
State	Published - 2018
Event	10th USENIX Workshop on Hot Topics in Storage and File Systems, HotStorage 2018, co-located with USENIX ATC 2018 - Boston, United States Duration: Jul 9 2018 → Jul 10 2018

Conference

Conference	10th USENIX Workshop on Hot Topics in Storage and File Systems, HotStorage 2018, co-located with USENIX ATC 2018
Country/Territory	United States
City	Boston
Period	7/9/18 → 7/10/18

ASJC Scopus subject areas

Hardware and Architecture
Information Systems
Software
Computer Networks and Communications

Cite this

@conference{8a1cd314711e4589a446a2581dd6ffd2,

title = "Driving cache replacement with ML-based LeCaR",

abstract = "Can machine learning (ML) be used to improve on existing cache replacement strategies? We propose a general framework called LeCaR that uses the ML technique of regret minimization to answer the question in the affirmative. We show that the LeCaR framework outperforms ARC using only two fundamental eviction policies, LRU and LFU, by more than 18x when the cache size is small relative to the size of the working set.",

author = "Giuseppe Vietri and Rodriguez, {Liana V.} and Martinez, {Wendy A.} and Steven Lyons and Jason Liu and Raju Rangaswami and Ming Zhao and Giri Narasimhan",

note = "Funding Information: We thanks the anonymous reviewers and our shepherd, Ioan Stefanovici, for their insightful feedback. This work is supported in part by NSF awards, CCF-1718335, CNS-1563883 and CNS-1562837, and a NetApp Faculty Fellowship. Publisher Copyright: {\textcopyright} USENIX Workshop on Hot Topics in Storage and File Systems, HotStorage 2018, co-located with USENIX ATC 2018.All right reserved.; 10th USENIX Workshop on Hot Topics in Storage and File Systems, HotStorage 2018, co-located with USENIX ATC 2018 ; Conference date: 09-07-2018 Through 10-07-2018",

year = "2018",

language = "English (US)",

}

TY - CONF

T1 - Driving cache replacement with ML-based LeCaR

AU - Vietri, Giuseppe

AU - Rodriguez, Liana V.

AU - Martinez, Wendy A.

AU - Lyons, Steven

AU - Liu, Jason

AU - Rangaswami, Raju

AU - Zhao, Ming

AU - Narasimhan, Giri

N1 - Funding Information: We thanks the anonymous reviewers and our shepherd, Ioan Stefanovici, for their insightful feedback. This work is supported in part by NSF awards, CCF-1718335, CNS-1563883 and CNS-1562837, and a NetApp Faculty Fellowship. Publisher Copyright: © USENIX Workshop on Hot Topics in Storage and File Systems, HotStorage 2018, co-located with USENIX ATC 2018.All right reserved.

PY - 2018

Y1 - 2018

N2 - Can machine learning (ML) be used to improve on existing cache replacement strategies? We propose a general framework called LeCaR that uses the ML technique of regret minimization to answer the question in the affirmative. We show that the LeCaR framework outperforms ARC using only two fundamental eviction policies, LRU and LFU, by more than 18x when the cache size is small relative to the size of the working set.

AB - Can machine learning (ML) be used to improve on existing cache replacement strategies? We propose a general framework called LeCaR that uses the ML technique of regret minimization to answer the question in the affirmative. We show that the LeCaR framework outperforms ARC using only two fundamental eviction policies, LRU and LFU, by more than 18x when the cache size is small relative to the size of the working set.

UR - http://www.scopus.com/inward/record.url?scp=85088225389&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85088225389&partnerID=8YFLogxK

M3 - Paper

AN - SCOPUS:85088225389

T2 - 10th USENIX Workshop on Hot Topics in Storage and File Systems, HotStorage 2018, co-located with USENIX ATC 2018

Y2 - 9 July 2018 through 10 July 2018

ER -

Driving cache replacement with ML-based LeCaR

Abstract

Conference

ASJC Scopus subject areas

Other files and links

Fingerprint

Cite this