TY - GEN
T1 - Assise: Performance and Availability via Client-local NVM in a Distributed File System
AU - Anderson, Thomas E.
AU - Canini, Marco
AU - Kim, Jongyul
AU - Kostic, Dejan
AU - Kwon, Youngjin
AU - Peter, Simon
AU - Reda, Waleed
AU - Schuh, Henry N.
AU - Witchel, Emmett
N1 - KAUST Repository Item: Exported on 2021-08-12
Acknowledgements: Waleed Reda was supported by a fellowship from the Erasmus Mundus Joint Doctorate in Distributed Computing (EMJD-DC), funded by the European Commission (EACEA) (FPA 2012-0030). This work is supported in part by ERC grant 770889, NSF grant CNS-1900457, and the Texas Systems Research Consortium. This work is also supported by the National Research Foundation of Korea (NRF) grant funded by the Korea government (MSIT) (2020R1C1C1014940). We thank Intel for access to the evaluation testbed. We thank the anonymous reviewers and our shepherd, Kim Keeton, for their comments and feedback.
PY - 2020
Y1 - 2020
N2 - he adoption of low latency persistent memory modules (PMMs) upends the long-established model of remote storage for distributed file systems. Instead, by colocating computation with PMM storage, we can provide applications with much higher IO performance, sub-second application failover, and strong consistency. To demonstrate this, we built the Assise distributed file system, based on a persistent, replicated coherence protocol that manages client-local PMM as a linearizable and crash-recoverable cache between applications and slower (and possibly remote) storage. Assise maximizes locality for all file IO by carrying out IO on process-local, socket-local, and client-local PMM whenever possible. Assise minimizes coherence overhead by maintaining consistency at IO operation granularity, rather than at fixed block sizes.
We compare Assise to Ceph/BlueStore, NFS, and Octopus on a cluster with Intel Optane DC PMMs and SSDs for common cloud applications and benchmarks, such as LevelDB, Postfix, and FileBench. We find that Assise improves write latency up to 22x, throughput up to 56x, fail-over time up to 103x, and scales up to 6x better than its counterparts, while providing stronger consistency semantics.
AB - he adoption of low latency persistent memory modules (PMMs) upends the long-established model of remote storage for distributed file systems. Instead, by colocating computation with PMM storage, we can provide applications with much higher IO performance, sub-second application failover, and strong consistency. To demonstrate this, we built the Assise distributed file system, based on a persistent, replicated coherence protocol that manages client-local PMM as a linearizable and crash-recoverable cache between applications and slower (and possibly remote) storage. Assise maximizes locality for all file IO by carrying out IO on process-local, socket-local, and client-local PMM whenever possible. Assise minimizes coherence overhead by maintaining consistency at IO operation granularity, rather than at fixed block sizes.
We compare Assise to Ceph/BlueStore, NFS, and Octopus on a cluster with Intel Optane DC PMMs and SSDs for common cloud applications and benchmarks, such as LevelDB, Postfix, and FileBench. We find that Assise improves write latency up to 22x, throughput up to 56x, fail-over time up to 103x, and scales up to 6x better than its counterparts, while providing stronger consistency semantics.
UR - http://hdl.handle.net/10754/670570
UR - https://www.usenix.org/conference/osdi20/presentation/anderson
M3 - Conference contribution
SP - 1011
EP - 1027
BT - Proceedings of the 14th USENIX Symposium on Operating Systems Design and Implementation
ER -