Policy-based management has been proposed in recent years as a suitable means for managing Quality of Service (QoS) in IP networks. Yet despite research projects, standardisation efforts, and substantial interest from industry, policy-based management is still not a reality. One of the reasons for the reticence to adopt this technology is that it is difficult to analyse policies to determine that they will actually work, given the capabilities of managed network devices, and to guarantee the stability of the network configuration, given that policies may have conflicts leading to unpredictable effects. This project aims to address the challenges of policy analysis, policy validation and policy refinement within the specific application domain of Quality of Service for IP networks. Policy analysis includes the ability to check for conflicts in the presence of constraints on the state of the system and identify precisely the sequence of events which would lead to a conflict as well as the ability to check that certain properties are satisfied when a given set of policies is applied. Policy validation relates to the ability to determine whether a given set of policies can actually be implemented on a network configuration given the characteristics of the devices and the traffic profiles. Policy refinement represents the ability to derive lower-level more concrete policies from higher level goals expressed in organisational policies and service level agreements (SLA). Policy refinement cannot be fully automated and the problem is extremely difficult to solve in the general case. We will develop an approach to partially automate this process through the identification and instantiation of domain specific refinement patterns of management policies, which will probably depend on the characteristics of the network devices and of the network itself. More specifically the project is targeted at Differentiated Services (DiffServ) in MPLS environments. In addition we will investigate other application domains such as security and ubiquitous e-health in related projects.To achieve these goals it is necessary to combine formal reasoning techniques with policy-based management approaches and domain specific knowledge on QoS in IP networks. In particular this project aims to provide a holistic approach by combining the analysis and refinement aspects with the design, extension and improvement of the QoS framework developed at the University of Surrey. This will not only provide a framework in which the results of the project can be validated through actual implementation on a testbed but also a framework where it will be possible to explore the limits of policy-based programmability of networks. The project will identify which parts of the QoS management framework can be implemented through a policy-based control and which parts and algorithms are better provided through intelligence within managed objects.
This project was in collaboration with Prof. George Pavlou at the University of Surrey (now at University College London).