Multi-agent Reinforcement Learning in Network Management
AIMS '09 Proceedings of the 3rd International Conference on Autonomous Infrastructure, Management and Security: Scalability of Networks and Services
Hi-index | 0.00 |
This paper presents a policy-based architecture for adaptable service systems based on the combination of Reasoning Machines and Extended Finite State Machines. Policies are introduced to obtain flexibility with respect to specification and execution of adaptable service systems that give high performance over a range of system status values. The presented architecture covers three aspects: service system framework, adaptation mechanisms and data model. The adaptation mechanisms can be based on static or dynamic policy systems. Static policy systems have a non-changeable set of policies, Dynamic policy systems have a changeable set of policies, which are managed by policies at a higher level. The data model for the reasoning machine functionality is based on the rule-based reasoning language "XML Equivalent Transformation" (XET). The capability configuration mangement of a service system with runtime simulation results based on the proposed architecture is presented with the intention to illustrate the use of the architecture and discuss the potential advantages of using dynamic policies.