| 摘要 |
To control the load on an agent server, it is necessary to store any inactive agent in a secondary storage device, release a memory and a thread occupied by the agent, and control the number of simultaneously active agents. The agent server reads out an agent from the secondary storage device when a processing of the agent starts, and writes the agent into the secondary storage device when the processing of the agent is terminated. By doing so, the memory and the thread occupied by the agents which are not processing are released. It also limits the number of simultaneously active agents by starting a processing of another agent upon termination of the processing of one agent. In the present invention, the start and the termination of the processing are determined by the start and the termination of a conversation between the agents. Messages are in compliance with a predetermined rule. Accordingly, if this conversation rule includes definition of types of messages to the flow of messages exchanged in a series of processes, the start and the termination of a conversation can be detected by referring to the message and the conversation rule.
|
