DYNAMIC LOAD CONTROL USING A MULTISERVER CLIENT MODEL-BASED ON DISTRIBUTED AND AUTONOMOUS AGENTS

Seamless connection between LAN and WAN using high speed networks call provide real time global multi-me dia serves environments, such as au dio and video, to an unspecified number of clients. In this type of se rver-client environment it is necessary to select servers appropriatel y and in real time, based on the load condition of the server, the ser vice, and the quality of service (QoS) requirements of clients. For th is kind of system, a new concept that can achieve autonomous cooperati on among the servers is needed. It should be able to learn from the re sults and then re-establish a relationship with the clients based on p redicted environmental fluctuations. In this paper we propose a dynami c load control system with autonomous cooperation and time domain pred iction capabilities based on a multi-server client model having multi- media load control and implemented as an autonomous distributed agent system. Furthermore, CLOS is used to install an evaluation system and tile adaptability is evaluated. It is confirmed that by learning the t ime dependence of the load fluctuation pattern, a prediction capabilit y can be obtained within a small number of learning cycles. (C) 1998 S cripta Technica.