Database Engines on Multicores Scale:  A Practical Approach

João Soares; Nuno Preguiça

Database Engines on Multicores Scale: A Practical Approach

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

Abstract

Multicore processors are available for over a decade, being the norm for current computer systems, but general purpose database management systems (DBMS) still cannot fully explore the computational resources of these platforms.
We focus on In-Memory DBMS since these are becoming widely adopted, due to the increasing amount of memory installed in today’s systems, and are expected to scale on multicore machines, by not incurring in I/O bottlenecks.
This paper presents a practical study on In-Memory DBMS and shows that contention imposed by concurrency control mechanisms, such as locking, are limiting factors for both performance and scalability of these systems on multicores.
Additionally, we discuss a simple database engine modifica-
tion that allows an almost 10 fold performance improvement, over the original engine, also allowing databases to scale on multicores.

Original language	English
Title of host publication	Proceedings of the 30th Annual ACM Symposium on Applied Computing
Place of Publication	New York, NY, USA
Publisher	ACM - Association for Computing Machinery
Pages	1297–1302
ISBN (Print)	978-1-4503-3196-8
Publication status	Published - 2015

Publication series

Name	SAC '15
Publisher	ACM

Keywords

cache
concurrency control
distributed systems
multi-version
partial data replication

Cite this

@inproceedings{3c8acbb8c80a46bfaff410a8dab7e5b2,

title = "Database Engines on Multicores Scale: A Practical Approach",

abstract = "Multicore processors are available for over a decade, being the norm for current computer systems, but general purpose database management systems (DBMS) still cannot fully explore the computational resources of these platforms.We focus on In-Memory DBMS since these are becoming widely adopted, due to the increasing amount of memory installed in today{\textquoteright}s systems, and are expected to scale on multicore machines, by not incurring in I/O bottlenecks.This paper presents a practical study on In-Memory DBMS and shows that contention imposed by concurrency control mechanisms, such as locking, are limiting factors for both performance and scalability of these systems on multicores.Additionally, we discuss a simple database engine modifica-tion that allows an almost 10 fold performance improvement, over the original engine, also allowing databases to scale on multicores.",

keywords = "cache , concurrency control, distributed systems, multi-version, partial data replication",

author = "Jo{\~a}o Soares and Nuno Pregui{\c c}a",

note = "sem pdf",

year = "2015",

language = "English",

isbn = "978-1-4503-3196-8",

series = "SAC '15",

publisher = "ACM - Association for Computing Machinery",

pages = "1297–1302",

booktitle = "Proceedings of the 30th Annual ACM Symposium on Applied Computing",

address = "United States",

}

TY - GEN

T1 - Database Engines on Multicores Scale

T2 - A Practical Approach

AU - Soares, João

AU - Preguiça, Nuno

N1 - sem pdf

PY - 2015

Y1 - 2015

N2 - Multicore processors are available for over a decade, being the norm for current computer systems, but general purpose database management systems (DBMS) still cannot fully explore the computational resources of these platforms.We focus on In-Memory DBMS since these are becoming widely adopted, due to the increasing amount of memory installed in today’s systems, and are expected to scale on multicore machines, by not incurring in I/O bottlenecks.This paper presents a practical study on In-Memory DBMS and shows that contention imposed by concurrency control mechanisms, such as locking, are limiting factors for both performance and scalability of these systems on multicores.Additionally, we discuss a simple database engine modifica-tion that allows an almost 10 fold performance improvement, over the original engine, also allowing databases to scale on multicores.

AB - Multicore processors are available for over a decade, being the norm for current computer systems, but general purpose database management systems (DBMS) still cannot fully explore the computational resources of these platforms.We focus on In-Memory DBMS since these are becoming widely adopted, due to the increasing amount of memory installed in today’s systems, and are expected to scale on multicore machines, by not incurring in I/O bottlenecks.This paper presents a practical study on In-Memory DBMS and shows that contention imposed by concurrency control mechanisms, such as locking, are limiting factors for both performance and scalability of these systems on multicores.Additionally, we discuss a simple database engine modifica-tion that allows an almost 10 fold performance improvement, over the original engine, also allowing databases to scale on multicores.

KW - cache

KW - concurrency control

KW - distributed systems

KW - multi-version

KW - partial data replication

M3 - Conference contribution

SN - 978-1-4503-3196-8

T3 - SAC '15

SP - 1297

EP - 1302

BT - Proceedings of the 30th Annual ACM Symposium on Applied Computing

PB - ACM - Association for Computing Machinery

CY - New York, NY, USA

ER -

Database Engines on Multicores Scale: A Practical Approach

Abstract

Publication series

Keywords

Fingerprint

Cite this