2nd International IEEE Conference on Communication System Software and Middleware

Research Article

A Fair and Effective Scheduling Algorithm for Multi-Agent Systems

  • @INPROCEEDINGS{10.1109/COMSWA.2007.382583,
        author={Vijitha Ranatunga and Akira Hosokawa and Kazuhiko Kinoshita and Nariyoshi  Yamait and Koso Murakami},
        title={A Fair and Effective Scheduling Algorithm for Multi-Agent Systems},
        proceedings={2nd International IEEE Conference on Communication System Software and Middleware},
        publisher={IEEE},
        proceedings_a={COMSWARE},
        year={2007},
        month={7},
        keywords={Computational modeling  Computer networks  Delay  Distributed computing  Grid computing  Information science  Mobile agents  Multiagent systems  Processor scheduling  Scheduling algorithm},
        doi={10.1109/COMSWA.2007.382583}
    }
    
  • Vijitha Ranatunga
    Akira Hosokawa
    Kazuhiko Kinoshita
    Nariyoshi Yamait
    Koso Murakami
    Year: 2007
    A Fair and Effective Scheduling Algorithm for Multi-Agent Systems
    COMSWARE
    IEEE
    DOI: 10.1109/COMSWA.2007.382583
Vijitha Ranatunga1,*, Akira Hosokawa1,*, Kazuhiko Kinoshita1,*, Nariyoshi Yamait2,*, Koso Murakami1,*
  • 1: Department of Information Networking, Graduate School of Information Science and Technology, Osaka University 1-5 Yamadaoka, Suita, Osaka 565-0871, Japan
  • 2: lnformation Technology Center, Okayama University 3-1-1 Tsushimanaka, Okayama, Okayama 700-8530, Japan
*Contact email: vijitha.ranatung@ist.osaka-u.ac.jp, akira@ist.osaka-u.ac.jp, kazuhiko@ist.osaka-u.ac.jp, yamai@cc.okayama-u.ac.jp, murakami@ist.osaka-u.ac.jp

Abstract

With the rapid increase of global connectivity, various types of distributed computer systems are being implemented to process information located in all parts of the world. Some distributed computing systems such as multi-agent systems require the same task to be applied to many individual instances of data located on disparate nodes. In such an environment, however, some agents may finish earlier than the others, due to the difference in the volume of data to be processed on a node and/or competition with other users' agents running on the same node. In this case, the time taken to complete the task depends on the slowest agent. Most of the existing scheduling algorithms hardly consider the status of the whole system for this type of application. In this paper, we propose and evaluate a fair and effective scheduling algorithm for multi-agent systems based on the processing status of each agent, the CPU power of each node, and the computational demand of each user. Simulation experiments show excellent performance of our scheduling algorithm.