Deepening the theory of queueing systems with variable processing rate and their applications in energy-saving data-centers

2021 Fiscal Year Final Research Report

• Project/Area Number: 18K18006
• Research Category: Grant-in-Aid for Early-Career Scientists
• Allocation Type: Multi-year Fund
• Review Section: Basic Section 60020:Mathematical informatics-related
• Research Institution: University of Tsukuba
• Principal Investigator: Phung-Duc Tuan 筑波大学, システム情報系, 准教授 (20633465)
• Project Period (FY): 2018-04-01 – 2022-03-31
• Keywords: マルコフ連鎖 / 待ち行列理論 / 流体極限 / 拡散極限 / データセンター / スケール極限 / 省エネ

Outline of Final Research Achievements

Large-scale data centers are important information infrastructure that supports the modern information society. A large data center consists of tens of thousands of computers and consumes a large amount of power. However, in reality, the rate that these computers are busy processing jobs remains at 30-40%. Therefore, it is expected that the job delay time will be reduced while reducing the power consumption by turning off non-operating server and restarting the stopped server when the load increases. In this research, we propose the appropriate operation policy of the data center and evaluate the performance by modeling the relevant data center by queueing systems with a variable processing speed and analyzing the performance.

Free Research Field

数理情報学

Academic Significance and Societal Importance of the Research Achievements

昨今の情報社会において，Facebookなどのソーシャルネットワーク，Zoom，MS Teamsなどのビデオ会議，そしてDropboxなどのストレージサービスは日常生活及び仕事で不可欠となっている．これらのシステムを支える情報インフラ基盤はデータセンターである．データセンターでは多数のコンピュータが存在して，大量の電力を消費している．本研究課題では大規模データセンターを数理モデル化し，モデルを解析することで，システムを効率的に運営するための性能評価量を提供する．