Derecho: Fast state machine replication for cloud services

Sagar Jha; Jonathan Behrens; Theo Gkountouvas; Matthew Milano; Weijia Song; Edward Tremel; Robbert Van Renesse; Sydney Zink; Kenneth P. Birman

doi:10.1145/3302258

Derecho: Fast state machine replication for cloud services

Sagar Jha, Jonathan Behrens, Theo Gkountouvas, Matthew Milano, Weijia Song, Edward Tremel, Robbert Van Renesse, Sydney Zink, Kenneth P. Birman

Research output: Contribution to journal › Article › peer-review

37 Scopus citations

Abstract

Cloud computing services often replicate data and may require ways to coordinate distributed actions. Here we present Derecho, a library for such tasks. The API provides interfaces for structuring applications into patterns of subgroups and shards, supports state machine replication within them, and includes mechanisms that assist in restart after failures. Running over 100Gbps RDMA, Derecho can send millions of events per second in each subgroup or shard and throughput peaks at 16GB/s, substantially outperforming prior solutions. Configured to run purely on TCP, Derecho is still substantially faster than comparable widely used, highly-tuned, standard tools. The key insight is that on modern hardware (including non-RDMA networks), data-intensive protocols should be built from non-blocking data-flow components.

Original language	English (US)
Article number	4
Journal	ACM Transactions on Computer Systems
Volume	36
Issue number	2
DOIs	https://doi.org/10.1145/3302258
State	Published - 2019
Externally published	Yes

Keywords

Cloud computing
Consistency
Non-volatile memory
RDMA
Replication

ASJC Scopus subject areas

General Computer Science

Access to Document

10.1145/3302258

Cite this

@article{42993bf3ec6545469d74e13aec08bff6,

title = "Derecho: Fast state machine replication for cloud services",

abstract = "Cloud computing services often replicate data and may require ways to coordinate distributed actions. Here we present Derecho, a library for such tasks. The API provides interfaces for structuring applications into patterns of subgroups and shards, supports state machine replication within them, and includes mechanisms that assist in restart after failures. Running over 100Gbps RDMA, Derecho can send millions of events per second in each subgroup or shard and throughput peaks at 16GB/s, substantially outperforming prior solutions. Configured to run purely on TCP, Derecho is still substantially faster than comparable widely used, highly-tuned, standard tools. The key insight is that on modern hardware (including non-RDMA networks), data-intensive protocols should be built from non-blocking data-flow components.",

keywords = "Cloud computing, Consistency, Non-volatile memory, RDMA, Replication",

author = "Sagar Jha and Jonathan Behrens and Theo Gkountouvas and Matthew Milano and Weijia Song and Edward Tremel and {Van Renesse}, Robbert and Sydney Zink and Birman, {Kenneth P.}",

note = "Publisher Copyright: {\textcopyright} 2019 Association for Computing Machinery.",

year = "2019",

doi = "10.1145/3302258",

language = "English (US)",

volume = "36",

journal = "ACM Transactions on Computer Systems",

issn = "0734-2071",

publisher = "Association for Computing Machinery (ACM)",

number = "2",

}

TY - JOUR

T1 - Derecho

T2 - Fast state machine replication for cloud services

AU - Jha, Sagar

AU - Behrens, Jonathan

AU - Gkountouvas, Theo

AU - Milano, Matthew

AU - Song, Weijia

AU - Tremel, Edward

AU - Van Renesse, Robbert

AU - Zink, Sydney

AU - Birman, Kenneth P.

PY - 2019

Y1 - 2019

N2 - Cloud computing services often replicate data and may require ways to coordinate distributed actions. Here we present Derecho, a library for such tasks. The API provides interfaces for structuring applications into patterns of subgroups and shards, supports state machine replication within them, and includes mechanisms that assist in restart after failures. Running over 100Gbps RDMA, Derecho can send millions of events per second in each subgroup or shard and throughput peaks at 16GB/s, substantially outperforming prior solutions. Configured to run purely on TCP, Derecho is still substantially faster than comparable widely used, highly-tuned, standard tools. The key insight is that on modern hardware (including non-RDMA networks), data-intensive protocols should be built from non-blocking data-flow components.

AB - Cloud computing services often replicate data and may require ways to coordinate distributed actions. Here we present Derecho, a library for such tasks. The API provides interfaces for structuring applications into patterns of subgroups and shards, supports state machine replication within them, and includes mechanisms that assist in restart after failures. Running over 100Gbps RDMA, Derecho can send millions of events per second in each subgroup or shard and throughput peaks at 16GB/s, substantially outperforming prior solutions. Configured to run purely on TCP, Derecho is still substantially faster than comparable widely used, highly-tuned, standard tools. The key insight is that on modern hardware (including non-RDMA networks), data-intensive protocols should be built from non-blocking data-flow components.

KW - Cloud computing

KW - Consistency

KW - Non-volatile memory

KW - RDMA

KW - Replication

UR - http://www.scopus.com/inward/record.url?scp=85065756265&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85065756265&partnerID=8YFLogxK

U2 - 10.1145/3302258

DO - 10.1145/3302258

M3 - Article

AN - SCOPUS:85065756265

SN - 0734-2071

VL - 36

JO - ACM Transactions on Computer Systems

JF - ACM Transactions on Computer Systems

IS - 2

M1 - 4

ER -

Derecho: Fast state machine replication for cloud services

Abstract

Keywords

ASJC Scopus subject areas

Access to Document

Other files and links

Fingerprint

Cite this