Design and Implementation of a Distributed-Agent-based Simulation for Hierarchical Service-Deployment

Large and heterogeneous networks are difficult to manage, particularly when a wide set of customized services are expected to be deployed in such networks. This is especially true when parts of the network are programmable, such as Network Processors for instance. To help automate the deployment task of such services, a new protocol has been proposed. This paper describes the implementation aspects of the simulation used to evaluate the complexity of the protocol. The paper presents the design and usage of an agent-based extension to the network simulation software tool ns-2, as well as the implementation of the deployment of services of three categories, i.e., a distributed web-cache service, a service providing quality-of-service (QoS) using differentiated services (DiffServ), and a Virtual Private Network service providing QoS between multiple end-points, and also using DiffServ. A Straight Path Search algorithm is introduced, which reduces the amount of nodes queried, in order to perform optimal DiffServ deployment. In the VPN case, an adaptation of the Selective Closest Terminal First algorithm is presented that computes an approximation of the Minimum Steiner Tree over a hierarchical network.