FuseME: Distributed Matrix Computation Engine based on Cuboid-based Fused Operator and Plan Generation

Donghyoung Han, Jongwuk Lee, Min Soo Kim

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

1 Scopus citations

Abstract

Operator fusion is essentially and widely used in a large number of matrix computation systems in science and industry. The existing distributed operator fusion methods focus on only either low communication cost with the risk of out of memory or large-scale processing with high communication cost. We propose a distributed elastic fused operator called Cuboid-based Fused Operator (CFO) that achieves both low communication cost and large-scale processing. We also propose a novel fusion plan generator called Cuboid-based Fusion plan Generator (CFG) that finds a fusion plan to fuse more operators including large-scale matrix multiplication. We implement a fast distributed matrix computation engine called FuseME by integrating both CFO and CFG seamlessly. FuseME outperforms the state-of-the-art systems including SystemDS by orders of magnitude.

Original languageEnglish
Title of host publicationSIGMOD 2022 - Proceedings of the 2022 International Conference on Management of Data
PublisherAssociation for Computing Machinery
Pages1891-1904
Number of pages14
ISBN (Electronic)9781450392495
DOIs
StatePublished - Jun 2022
Event2022 ACM SIGMOD International Conference on the Management of Data, SIGMOD 2022 - Hybrid, Philadelphia, United States
Duration: 12 Jun 202217 Jun 2022

Publication series

NameProceedings of the ACM SIGMOD International Conference on Management of Data
ISSN (Print)0730-8078

Conference

Conference2022 ACM SIGMOD International Conference on the Management of Data, SIGMOD 2022
Country/TerritoryUnited States
CityHybrid, Philadelphia
Period12/06/2217/06/22

Keywords

  • distributed data-parallel system
  • matrix operators
  • operator fusion

Fingerprint

Dive into the research topics of 'FuseME: Distributed Matrix Computation Engine based on Cuboid-based Fused Operator and Plan Generation'. Together they form a unique fingerprint.

Cite this