TY - GEN
T1 - Strong and efficient consistency with consistency-aware durability
AU - Ganesan, Aishwarya
AU - Alagappan, Ramnatthan
AU - Arpaci-Dusseau, Andrea C.
AU - Arpaci-Dusseau, Remzi H.
N1 - Funding Information:
We thank Yu Hua (our shepherd) and the anonymous reviewers of FAST ’20 for their insightful comments and suggestions. We thank the members of ADSL for their excellent feedback. We also thank CloudLab [47] for providing a great environment to run our experiments. This material was supported by funding from NSF grants CNS-1421033, CNS-1763810 and CNS-1838733, and DOE grant DE-SC0014935. Aishwarya Ganesan is supported by a Facebook fellowship. Any opinions, findings, and conclusions or recommendations expressed in this material are those of the authors and may not reflect the views of NSF, DOE, or any other institutions.
Publisher Copyright:
Copyright © Proc. of the 18th USENIX Conference on File and Storage Tech., FAST 2020. All rights reserved.
PY - 2020
Y1 - 2020
N2 - We introduce consistency-aware durability or CAD, a new approach to durability in distributed storage that enables strong consistency while delivering high performance. We demonstrate the efficacy of this approach by designing cross-client monotonic reads, a novel and strong consistency property that provides monotonic reads across failures and sessions in leader-based systems. We build ORCA, a modified version of ZooKeeper that implements CAD and cross-client monotonic reads. We experimentally show that ORCA provides strong consistency while closely matching the performance of weakly consistent ZooKeeper. Compared to strongly consistent ZooKeeper, ORCA provides significantly higher throughput (1.8 - 3.3×), and notably reduces latency, sometimes by an order of magnitude in geo-distributed settings.
AB - We introduce consistency-aware durability or CAD, a new approach to durability in distributed storage that enables strong consistency while delivering high performance. We demonstrate the efficacy of this approach by designing cross-client monotonic reads, a novel and strong consistency property that provides monotonic reads across failures and sessions in leader-based systems. We build ORCA, a modified version of ZooKeeper that implements CAD and cross-client monotonic reads. We experimentally show that ORCA provides strong consistency while closely matching the performance of weakly consistent ZooKeeper. Compared to strongly consistent ZooKeeper, ORCA provides significantly higher throughput (1.8 - 3.3×), and notably reduces latency, sometimes by an order of magnitude in geo-distributed settings.
UR - http://www.scopus.com/inward/record.url?scp=85091819502&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85091819502&partnerID=8YFLogxK
M3 - Conference contribution
AN - SCOPUS:85091819502
SP - 323
EP - 337
BT - Proceedings of the 18th USENIX Conference on File and Storage Technologies, FAST 2020
PB - USENIX Association
T2 - 18th USENIX Conference on File and Storage Technologies, FAST 2020
Y2 - 25 February 2020 through 27 February 2020
ER -