Declarative Recursive Computation on an RDBMS or, Why You Should Use a Database For Distributed Machine Learning

Dimitrije Jankov; Shangyu Luo; Binhang Yuan; Zhuhua Cai; Jia Zou; Chris Jermaine; Zekai J. Gao

doi:10.14778/3317315.3317323

Declarative Recursive Computation on an RDBMS or, Why You Should Use a Database For Distributed Machine Learning

Dimitrije Jankov, Shangyu Luo, Binhang Yuan, Zhuhua Cai, Jia Zou, Chris Jermaine, Zekai J. Gao

Research output: Contribution to journal › Conference article › peer-review

26 Scopus citations

Abstract

A number of popular systems, most notably Google's TensorFlow, have been implemented from the ground up to support machine learning tasks. We consider how to make a very small set of changes to a modern relational database management system (RDBMS) to make it suitable for distributed learning computations. Changes include adding better support for recursion, and optimization and execution of very large compute plans. We also show that there are key advantages to using an RDBMS as a machine learning platform. In particular, learning based on a database management system allows for trivial scaling to large data sets and especially large models, where different computational units operate on different parts of a model that may be too large to fit into RAM.

Original language	English (US)
Pages (from-to)	822-835
Number of pages	14
Journal	Proceedings of the VLDB Endowment
Volume	12
Issue number	7
DOIs	https://doi.org/10.14778/3317315.3317323
State	Published - 2018
Externally published	Yes
Event	45th International Conference on Very Large Data Bases, VLDB 2019 - Los Angeles, United States Duration: Aug 26 2017 → Aug 30 2017

ASJC Scopus subject areas

Computer Science (miscellaneous)
General Computer Science

Access to Document

10.14778/3317315.3317323

Cite this

@article{ebfe43455be64939983f03ba7e032125,

title = "Declarative Recursive Computation on an RDBMS or, Why You Should Use a Database For Distributed Machine Learning",

abstract = "A number of popular systems, most notably Google's TensorFlow, have been implemented from the ground up to support machine learning tasks. We consider how to make a very small set of changes to a modern relational database management system (RDBMS) to make it suitable for distributed learning computations. Changes include adding better support for recursion, and optimization and execution of very large compute plans. We also show that there are key advantages to using an RDBMS as a machine learning platform. In particular, learning based on a database management system allows for trivial scaling to large data sets and especially large models, where different computational units operate on different parts of a model that may be too large to fit into RAM.",

author = "Dimitrije Jankov and Shangyu Luo and Binhang Yuan and Zhuhua Cai and Jia Zou and Chris Jermaine and Gao, {Zekai J.}",

note = "Publisher Copyright: {\textcopyright} 2019, VLDB Endowment.; 45th International Conference on Very Large Data Bases, VLDB 2019 ; Conference date: 26-08-2017 Through 30-08-2017",

year = "2018",

doi = "10.14778/3317315.3317323",

language = "English (US)",

volume = "12",

pages = "822--835",

journal = "Proceedings of the VLDB Endowment",

issn = "2150-8097",

publisher = "Very Large Data Base Endowment Inc.",

number = "7",

}

TY - JOUR

T1 - Declarative Recursive Computation on an RDBMS or, Why You Should Use a Database For Distributed Machine Learning

AU - Jankov, Dimitrije

AU - Luo, Shangyu

AU - Yuan, Binhang

AU - Cai, Zhuhua

AU - Zou, Jia

AU - Jermaine, Chris

AU - Gao, Zekai J.

PY - 2018

Y1 - 2018

N2 - A number of popular systems, most notably Google's TensorFlow, have been implemented from the ground up to support machine learning tasks. We consider how to make a very small set of changes to a modern relational database management system (RDBMS) to make it suitable for distributed learning computations. Changes include adding better support for recursion, and optimization and execution of very large compute plans. We also show that there are key advantages to using an RDBMS as a machine learning platform. In particular, learning based on a database management system allows for trivial scaling to large data sets and especially large models, where different computational units operate on different parts of a model that may be too large to fit into RAM.

AB - A number of popular systems, most notably Google's TensorFlow, have been implemented from the ground up to support machine learning tasks. We consider how to make a very small set of changes to a modern relational database management system (RDBMS) to make it suitable for distributed learning computations. Changes include adding better support for recursion, and optimization and execution of very large compute plans. We also show that there are key advantages to using an RDBMS as a machine learning platform. In particular, learning based on a database management system allows for trivial scaling to large data sets and especially large models, where different computational units operate on different parts of a model that may be too large to fit into RAM.

UR - http://www.scopus.com/inward/record.url?scp=85077239058&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85077239058&partnerID=8YFLogxK

U2 - 10.14778/3317315.3317323

DO - 10.14778/3317315.3317323

M3 - Conference article

AN - SCOPUS:85077239058

SN - 2150-8097

VL - 12

SP - 822

EP - 835

JO - Proceedings of the VLDB Endowment

JF - Proceedings of the VLDB Endowment

IS - 7

T2 - 45th International Conference on Very Large Data Bases, VLDB 2019

Y2 - 26 August 2017 through 30 August 2017

ER -

Declarative Recursive Computation on an RDBMS or, Why You Should Use a Database For Distributed Machine Learning

Abstract

ASJC Scopus subject areas

Access to Document

Other files and links

Fingerprint

Cite this