On the Design of an Architecture for Partitioned Knowledge Management in Autonomic Multimedia Access and Aggregation Networks Steven Latr´ e, Stijn Verstichel, Bart De Vleeschauwer, Filip De Turck, Piet Demeester IBCN, Department of Information Technology, Ghent University - IBBT Gaston Crommenlaan 8/201, 9050 Gent, Belgium Steven.Latre@intec.ugent.be Abstract. The recent emergence of multimedia services, such as Net- work Based Personal Video Recording and Broadcast TV over tradi- tional DSL based access networks, has introduced stringent Quality of Experience (QoE) requirements. It is generally assumed that the wide variety of services and user profiles introduces the need for a per-user or per-subscriber QoE management. Such a complex QoE management re- quires real-time knowledge about the managed services, which is available amongst the different nodes in the network. However, even for managing a few services, a relatively large amount of, constantly updated, knowl- edge is needed. Propagating all the knowledge to all nodes is therefore not feasible. As not all knowledge is relevant to all nodes, it is impor- tant to perform an intelligent knowledge distribution and management. In this position paper, we introduce the concept of a cognitive model that describes the knowledge requirements of each node. Based on the information stated in this cognitive model, we discuss how filter queries, that typically describe what needs to be queried from other nodes, can be automatically generated leading to an efficient partitioning of the knowledge through the distributed nodes. 1 Introduction Multimedia services over broadband DSL access and aggregation networks such as Broadcast TV and Video on Demand have gained a lot of popularity in the last few years. For end users, these services introduce new possibilities such as interactivity and higher video quality. For service operators, multimedia services offer an increased revenue in their Triple Play offer. At the same time, these multimedia services have stringent quality requirements: they often require a substantial amount of bandwidth and tolerate no packet loss and only small amounts of jitter. Operators who want to maximise their revenue try to manage the service quality as perceived by the end user, commonly described as the Quality of Experience (QoE). This QoE management is further complicated by the heterogeneity of today’s access and aggregation networks. A wide variety of service enablers such as proxies, caches, admission control techniques and