An agent-inspired active network resource trading model applied to congestion control

peer reviewedIn order to accommodate fluctuations in network conditions, adaptive applications need to obtain information about resource availability, Using active networks, new models for adaptive applications can be envisaged, which can benefit from the possibility to send mobile code to the network nodes. We describe a model for trading resources inside an active net-work node, based on the interaction between capsules as reactive user agents, and resource manager agents which reside in the network nodes. We apply the model to the case of a many-to-one audio application with congestion control, which trades off link resources against memory when there is congestion at the outgoing interface towards the destination. Our simulation results indicate that the application makes effective use of the available resources, and it also allows resources to be shared according to user preferences