Task Distribution/Assignment and Energy Efficiency Optimization on Virtual Environments by Utilizing Parallelizing Compiler Analysis Information of User Applications

2019 Fiscal Year Final Research Report

• Project/Area Number: 17K12665
• Research Category: Grant-in-Aid for Young Scientists (B)
• Allocation Type: Multi-year Fund
• Research Field: Software
• Research Institution: Meisei University
• Principal Investigator: WADA Yasutaka 明星大学, 情報学部, 准教授 (40434310)
• Project Period (FY): 2017-04-01 – 2020-03-31
• Keywords: 低消費電力化 / 性能モデリング / アプリケーション解析 / 仮想環境

Outline of Final Research Achievements

In order to improve the efficiency of a virtualized environment, which is the basis of today's cloud services, we have realized/proposed: a) a power-performance modeling strategy for parallel applications and its implementation on a software framework, b) a task sharing method between a small-size edge device and a large server for an efficient deep-learning processing (both of learning and inference), and c) a management method and interface for power-performance control requests from multiple virtual machines to control an actual hardware appropriately and safely. With these achievements and proposals, it is possible to make cloud systems more efficient easily.

Free Research Field

計算機工学

Academic Significance and Societal Importance of the Research Achievements

我々が日常的に利用するクラウドサービスを支えるコンピュータシステムはその規模・数を日々増大させており，サービスの質を保ちつつ消費電力を削減することは急務である．本研究課題はクラウドシステム，引いては並列システムや並列アプリケーションの高効率化を目指し，システムとアプリケーションの両面からそれを実現しようとするものである．その成果を発展・活用させることで，より利便性が高く，かつ持続可能性の高い社会の実現に資することができると考えられる．