The tremendous growth in ubiquity and diversity of wireless networks render the exploitation of highly heterogeneous next generation networks more challenging. This complexity is compounded by the highly dynamic nature of many of these network types and exacerbates the demands on the network management system to ensure robust service delivery. Various centralised, distributed or cooperative management systems have been proposed to address these demands. However, an inability to scale in some cases, or weak support for automation and lack of control for operators in others render existing approaches unsuitable for managing networks in a cost-effective way. For this reason, we introduce Omega, a distributed and policy-based network management system with self-configuration capabilities. Every self-ware entity (i.e. network element) in the system is potentially a manager for a set of local (re)configuration tasks or for a sector of the network. The system may dynamically delegate management tasks that are beyond its remit (but necessary in order to achieve its goals) to other entities, re-evaluating this delegation of tasks according to changing topology and connectivity. The scalability of our platform comes from both (i) the minimisation of the traffic generated by management messages (e.g. events and configuration requests) and (ii) the use of a fully distributed discovery mechanism. Similarly, its adaptability is attributed to its capacity to react dynamically and in a timely way to network changes. This capability is provided by a resource discovery protocol that maintains an up-to-date overview of the network and a policy-based manager that exploits static knowledge about network configuration constraints and procedures, learnt information about network behaviour patterns and operators' strategic business objectives to trigger, execute and adapt configuration tasks to current network conditions.