Journal Article 動的エージェント負荷分散機構の開発
Development of the Dynamically Agent Load Balancer for a Multi Agent System

村田, 悠也  ,  山本, 学  ,  寺野, 隆雄

58 ( 3 )  , pp.754 - 761 , 2017-03-15
This paper discusses a large scale data aggregation problem implemented multi-agent system techniques. The multi-agent system is an abstract programming model with processes and data, To implement such a system, we must use so many software agents assigned to multiprocessors. Each agent, which has functions of data records, message handlers, and processing logics, works message-passing with asynchronous and parallel manners. The system usually shows poor performances, because the processing resources tend to only small number of specific agents, Furthermore, such processes are dynamically according to the environmental conditions, thus, it is difficult to correctly estimate the performance. To cope with the problem, this paper proposes a novel performance balancing method, which dynamically increases and/or decreases the number of agents against the changes of required computational resources. The proposed method enables us to balance the performance loads to automatically measure the dynamic performance by agents themselves. We apply the proposed method to a data aggregation application system, which process five hundred thousand per second. The experimental results show the 1.6 times processing improvement compared with a conventional method.

Number of accesses :  

Other information