Rein: Taming Tail Latency in Key-Value Stores via Multiget Scheduling

Handle URI:
http://hdl.handle.net/10754/625020
Title:
Rein: Taming Tail Latency in Key-Value Stores via Multiget Scheduling
Authors:
Reda, Waleed; Canini, Marco ( 0000-0002-5051-4283 ) ; Suresh, Lalith; Kostić, Dejan; Braithwaite, Sean
Abstract:
We tackle the problem of reducing tail latencies in distributed key-value stores, such as the popular Cassandra database.We focus on workloads of multiget requests, which batch together access to several data elements and parallelize read operations across the data store machines. We first analyze a production trace of a real system and quantify the skew due to multiget sizes, key popularity, and other factors. We then proceed to identify opportunities for reduction of tail latencies by recognizing the composition of aggregate requests and by carefully scheduling bottleneck operations that can otherwise create excessive queues. We design and implement a system called Rein, which reduces latency via inter-multiget scheduling using low overhead techniques. We extensively evaluate Rein via experiments in Amazon Web Services (AWS) and simulations. Our scheduling algorithms reduce the median, 95, and 99 percentile latencies by factors of 1.5, 1.5, and 1.9, respectively.
KAUST Department:
KAUST
Citation:
Reda W, Canini M, Suresh L, Kostić D, Braithwaite S (2017) Rein. Proceedings of the Twelfth European Conference on Computer Systems - EuroSys ’17. Available: http://dx.doi.org/10.1145/3064176.3064209.
Publisher:
ACM Press
Journal:
Proceedings of the Twelfth European Conference on Computer Systems - EuroSys '17
Conference/Event name:
12th European Conference on Computer Systems, EuroSys 2017
Issue Date:
17-Apr-2017
DOI:
10.1145/3064176.3064209
Type:
Conference Paper
Sponsors:
Waleed Reda was supported by a fellowship from the Erasmus Mundus Joint Doctorate in Distributed Computing (EMJD-DC) program funded by the European Commission (EACEA) (FPA 2012-0030). This project is in part financially supported by the Swedish Foundation for Strategic Research.
Additional Links:
http://dl.acm.org/citation.cfm?doid=3064176.3064209
Appears in Collections:
Conference Papers

Full metadata record

DC FieldValue Language
dc.contributor.authorReda, Waleeden
dc.contributor.authorCanini, Marcoen
dc.contributor.authorSuresh, Lalithen
dc.contributor.authorKostić, Dejanen
dc.contributor.authorBraithwaite, Seanen
dc.date.accessioned2017-06-14T12:17:34Z-
dc.date.available2017-06-14T12:17:34Z-
dc.date.issued2017-04-17en
dc.identifier.citationReda W, Canini M, Suresh L, Kostić D, Braithwaite S (2017) Rein. Proceedings of the Twelfth European Conference on Computer Systems - EuroSys ’17. Available: http://dx.doi.org/10.1145/3064176.3064209.en
dc.identifier.doi10.1145/3064176.3064209en
dc.identifier.urihttp://hdl.handle.net/10754/625020-
dc.description.abstractWe tackle the problem of reducing tail latencies in distributed key-value stores, such as the popular Cassandra database.We focus on workloads of multiget requests, which batch together access to several data elements and parallelize read operations across the data store machines. We first analyze a production trace of a real system and quantify the skew due to multiget sizes, key popularity, and other factors. We then proceed to identify opportunities for reduction of tail latencies by recognizing the composition of aggregate requests and by carefully scheduling bottleneck operations that can otherwise create excessive queues. We design and implement a system called Rein, which reduces latency via inter-multiget scheduling using low overhead techniques. We extensively evaluate Rein via experiments in Amazon Web Services (AWS) and simulations. Our scheduling algorithms reduce the median, 95, and 99 percentile latencies by factors of 1.5, 1.5, and 1.9, respectively.en
dc.description.sponsorshipWaleed Reda was supported by a fellowship from the Erasmus Mundus Joint Doctorate in Distributed Computing (EMJD-DC) program funded by the European Commission (EACEA) (FPA 2012-0030). This project is in part financially supported by the Swedish Foundation for Strategic Research.en
dc.publisherACM Pressen
dc.relation.urlhttp://dl.acm.org/citation.cfm?doid=3064176.3064209en
dc.titleRein: Taming Tail Latency in Key-Value Stores via Multiget Schedulingen
dc.typeConference Paperen
dc.contributor.departmentKAUSTen
dc.identifier.journalProceedings of the Twelfth European Conference on Computer Systems - EuroSys '17en
dc.conference.date2017-04-23 to 2017-04-26en
dc.conference.name12th European Conference on Computer Systems, EuroSys 2017en
dc.conference.locationBelgrade, SRBen
dc.contributor.institutionUniversité catholique de Louvain and KTH Royal Institute of Technologyen
dc.contributor.institutionKTH Royal Institute of Technologyen
dc.contributor.institutionVMware Researchen
dc.contributor.institutionSoundClouden
kaust.authorCanini, Marcoen
All Items in KAUST are protected by copyright, with all rights reserved, unless otherwise indicated.