Developing a methodology for large-scale graph parallel processing based on program synthesis and transformation

2022 Fiscal Year Final Research Report

• Project/Area Number: 19K11901
• Research Category: Grant-in-Aid for Scientific Research (C)
• Allocation Type: Multi-year Fund
• Section: 一般
• Review Section: Basic Section 60050:Software-related
• Research Institution: Meiji University (2022); The University of Electro-Communications (2019-2021)
• Principal Investigator: Iwasaki Hideya 明治大学, 理工学部, 専任教授 (90203372)
• Co-Investigator(Kenkyū-buntansha): 江本 健斗 九州工業大学, 大学院情報工学研究院, 准教授 (00587470)
• Project Period (FY): 2019-04-01 – 2023-03-31
• Keywords: 大規模グラフ並列処理 / 頂点主体計算 / 部分グラフ主体計算 / プログラム運算

Outline of Final Research Achievements

The objective of this research was to develop a methodology both for efficiently developing programs for large-scale graph parallel/distributed processing and for executing programs on their suitable processing models. The following results were obtained. (1) As an adaptive co-existence mechanism between vertex-centric computation and subgraph-centric computation, we designed and developed a system that enables a program to be run initially in the vertex-centric manner and then in the subgraph-centric manner in the course of its computation. (2) We designed a domain-specific language (DSL) that can handle subsets of vertices at a higher level of abstraction than vertex-centric computation, and built a compiler for the DSL. (3) We implemented the basic part of a parallel and distributed processing system that enables both vertex- and subgraph-centric computation, and synchronous/asynchronous execution.

Free Research Field

プログラミング言語，並列処理

Academic Significance and Societal Importance of the Research Achievements

本研究は，大規模グラフ並列分散処理を，主として頂点主体と部分グラフ主体という代表的なふたつのモデル，さらに実行方式として同期的実行と非同期的実行に注目している．巨大なグラフデータに対する処理への需要が高まって久しいが，どのモデルや処理方式が適しているかは，問題そのものの性質，グラフの形状等に依存しており，実際に処理を行ってみないとわからないという側面がある．本研究による頂点主体計算と部分グラフ主体計算の適応的な共存機構，公平な実行基盤提供するシステム等は，大規模グラフ並列処理プログラムの効率的な開発に寄与するものである．