Work package 1: The context-aware Post-IP architecture

Workpackage chairLIP6
PartnersDevoteam, Netcenter, Ginkgo, PUC, Telecom SudParis, UFRJ, Unicamp

 The future post-IP architecture should be based on context-aware properties. So, the first work package will have the task to define the way to deploy a context-aware infrastructure. The platform to realize this functionality will be included in network elements. Physical and logical sensors (software entity, network components, and software agents) are used to collect context information related to the presence, the location, the identity and the profile of users and services. A typical context-aware software involves the localization of services and users, calling up services according to user behaviour, providing information for service composition, facilitating ad hoc communication mechanisms between users, and adapting the QoS to the changes in the environment. This project will explore two types of context aware infrastructures and what is the best way to introduce intelligence in the Horizon platform composed of a Knowledge plane and a
Piloting planes that will be finalized in work package 3.
More precisely, the Horizon platform will be provided by a multi-agent system to offer some intelligence. The multi-agent system is formed with agents situated in all network equipment (common to all virtual instances).
Agents will be based on Ginkgo technology but a large number of improvements have to be provided through the Horizon project. The Horizon agent is shown in Figure 7.

                                           Fig. 7. The Horizon agent

The different entities of the Horizon agent are as follows. Each agent maintains its own view of the network on the basis of information obtained through the knowledge plane. This agent-centric view of the network is called the Situated View, and is focusing on the agent's close network environment.
The Behaviours are autonomic software components permanently adapting themselves to the environment changes. Each of these Behaviours can be considered as a specialized function with some expert capabilities.
Each Behaviour is essentially a sense->decide->act loop. Typical categories of Behaviours are as follows:
• Producing knowledge for the Situated View in cooperation with other agents.
• Reasoning individually or collectively to evaluate the situation and decide to apply an appropriate action,
e.g. a Behaviour can simply be in charge of computing bandwidth availability on the network equipment (NE).
It can also regularly perform a complex diagnostic scenario or it can be dedicated to automatic recognition of specific network conditions.
• Acting onto the NE parameters, e.g. Behaviour can tune QoS parameters in a DiffServ context.
Behaviours have access to the Situated View which operates within each agent as a whiteboard shared among the agent's Behaviours.
The fourth entity, the rule engine, can be seen as a specific Behaviour that can or cannot be used depending on the memory space and real time constraints. The rule engine exploits the tolerance for imprecision and learning capabilities. At this juncture, the principal constituents are Fuzzy Logic, Neural Computing, Evolutionary Computation Machine Learning and Probabilistic Reasoning.
The activation, dynamic parameterization and scheduling of Behaviours (the rule engine is seen as a behaviour) within an agent is performed by the Dynamic Planner. The Dynamic Planner decides which Behaviours have to be active, when they have to be active and with which parameters. The Dynamic Planner detects changes in the Situated View and occurrence of external/internal events; from there, it pilots the reaction of the agent to changes in the network environment.

Task 1: State-of-the-art in context aware technologies. Task 1 is devoted to a state of the art in context aware technologies. This state of the art will provide either passive context-aware infrastructure or active context-aware infrastructure. Passive context-aware infrastructure: Context is raw information that, when correctly interpreted, identifies the characteristics of an entity. An entity can be a person, place, a device or any object that is relevant to the interaction between a user and the services. Context is a function of time and environment. The environment is in turn a function of the users, services, resources and other entities in the environment. In this phase the focus will be on the context gathering and representation. A data model and dissemination protocol will be developed to represent, store and manage context information. This includes classifying context sources, providing a unified context structural representation and developing mobile storage strategies with data replication techniques to insure the availability and the proximity of context information.
Active context-aware infrastructure introduces enriched context sources with techniques that will provide smart context information delivery. A context level agreement protocol will be explored to provide automatic context matching with the user’s profile, terminal capabilities and service requirements and offering. A particular attention will be devoted to the pro-active aspect of the smart context with an appropriate context dissemination protocol. The primary aim of the protocol is the adaptive distribution of context information among multiple mobile and fixed sources and destinations (e.g. devices, service components) using (negotiated) specific dissemination attribute such as power saving and cost. Context dissemination can be achieved in both pull and push modes.

Task 2: Choice of the context aware technology. This phase will provide the best choice of the context aware technology that should be implemented in the Horizon platform. The choice will be deduced from results of Task 1 and the expertise of Ginkgo in this field.

Task 3: Intelligence-oriented tools for networking. In this task, the work to be done will focus on intelligence-oriented tools that could provide intelligence in the network to decide about integration of different knowledge, solving local problems, deciding on the algorithms to be used. This intelligence should allow the network to react in real time to be able to feed the control algorithms to solve congestion situation or failure or any problem arising in the network. Some learning process could be added if necessary.
The goal of this task 3 is to develop the Horizon agent and the Behaviours (and rule engine) dedicated to bring some intelligence using AI technologies. This intelligence will be used in WP3 to realize the piloting system able to manage & control virtual networks, and the allocation of physical resources to the different virtual networks. The Dynamic Planner and the Behaviours dedicated to the piloting system (piloting algorithms) will be developed in WP3.

Post-IP the task Deadline Leader

Research areas