A DISCRETE-TIME LOAD BALANCING PROBLEM BY NEURO-DYNAMIC PROGRAMMING ALGORITHMS

• Inoie Atsushi

  Kanagawa Institute of Technology

• Ohno Katsuhisa

  Aichi Institute of Technology

Bibliographic Information

Other Title

• ニューロ・ダイナミックプログラミングによる負荷分散システムの離散時間負荷分散政策
• ニューロ ダイナミック プログラミング ニ ヨル フカ ブンサン システム ノ リサン ジカン フカ ブンサン セイサク

Abstract

The meaning of load balancing is to dispatch jobs among resources of a system for maximizing the system performance. This paper deals with a discrete-time optimal load balancing problem that minimizes an expected total cost. This problem is formulated as an undiscounted Markov decision process, and is solved by the modified policy iteration method. Since the modified policy iteration method can not solve practical sized problems due to the curse of dimensionality, a near-optimal load balancing policy is computed by neuro-dynamic programming algorithms. We further compare this policy with heuristic policies such as the random policy, round-robin policy and shortest policy.

Journal

• Transactions of the Operations Research Society of Japan

  Transactions of the Operations Research Society of Japan 49 (0), 46-61, 2006

  The Operations Research Society of Japan

Keywords

• 計算機
• ニューロ・ダイナミックプログラミング
• マルコフ決定過程
• シミュレーション
• 負荷分散システム

Details

• CRID: 1390282680733041792
• NII Article ID: 110006151984
• NII Book ID: AA11998080
• DOI: 10.15807/torsj.49.46
• ISSN: 21888280
• ISSN: 13498940
• NDL BIB ID: 8701754
• Web Site: http://id.ndl.go.jp/bib/8701754
• Web Site: https://ndlsearch.ndl.go.jp/books/R000000004-I8701754
• Web Site: https://www.jstage.jst.go.jp/article/torsj/49/0/49_KJ00004480119/_pdf
• Text Lang: ja
• Data Source:
  • JaLC
  • NDL
  • Crossref
  • CiNii Articles
  • KAKEN
• Abstract License Flag: Disallowed