SUMMARYIP networks, and particularly the Internet, were proposed to be a simple and robust support for homogeneous communications. This implies that only basic control mechanisms have to be performed by network elements. Communication management has to be performed by the terminals. However, the integration of new services and the increasing need for QoS require the network to be increasingly more fl exible and adaptive. New algorithms and protocols have been proposed by many research teams to address these issues, but these new algorithms tend to make network management and control more fl exible. Thus, manual confi guration of such fl exible and adaptive network architectures is very complex, if not impossible. Self-management is then a good opportunity to address this new complexity, and then to integrate more easily new services into the network. However, this self-management requires the equipment to carry much more knowledge and information than the actual control and management planes do. Global knowledge management schemes are therefore necessary to achieve this, including new policies for knowledge gathering, computing, sharing and providing. To address this particular need for knowledge management, several studies have proposed building a new plane, called the 'Knowledge Plane' (KP). This paper studies different propositions for this KP, and presents an original vision of what this KP should be. Our vision of the KP relies on the paradigm of situatedness. This paradigm was developed by research studies in the fi eld of multi-agent systems, which tend to solve complex problems using collaborative and autonomous agents (multi-agent technology has been largely described in Artifi cial Intelligence literature). These agents in our proposition are embedded within the network elements themselves. Their role is to share local and situated knowledge composing the global KP. We have also developed, as an illustration, a distributed intrusion detection system (IDS) based on the local IDS Snort.

No abstract

The development of the IEEE wireless technologies is promising the ultimate Internet service deployment on wireless and mobile infrastructures since they would offer larger bandwidth at cheaper price. However, it is disquieting to see that the different control algorithms supposed to be the heart of the wireless deployment are not evolving as fast as the wireless technologies do. Here we come up with the hard question: how to pilot these control algorithms? This paper is proposing a revolutionary tyre where all control algorithms are fed in a concerted manner by a self-steering system. In network terms, this paper is proposing a definitely new technology that will permit to optimize, secure, manage and control the wireless devices using an automatic pilot system.

Abstract. IP Networks, and particularly the Internet, were proposed to be a simple and robust support for heterogeneous communications. This implies that only basic controls have to be done by network elements. Connection management, along with transport, and more generally communication management, has to be done by the terminals. For example, error detection mechanisms, error recovery mechanisms with "Slow Start", are implemented within the transport protocol, managed by the terminals. However, integration of new services and increasing need for QoS require the network to be more and more flexible and adaptive. New algorithms and protocols are then proposed to address these issues, and include new configuration layers. Manual configuration of such network architectures is then very complex, if not impossible. We think that future core network elements will have to be more adaptive, but also more autonomic. Auto configuration is indeed a necessary condition to integrate new services in the network. We believe that auto configuration requires new knowledge provisioning and computing policies. This paper then presents an architecture of software agents, collaborative and autonomic. These agents are embedded inside the routers. Their role is to share local and situated knowledge, in order to control and optimize the existing control mechanism of the router.