Abstract : Tiny devices, smart objects, home gateways, sensor networks, have become an integral part of our everyday life and of our socio-economic ecosystem. Managing e ciently such environments is just as important as the devices themselves. However, new management challenges such as management in di erent scales (multi-scale), heterogeneity management or extreme managed resources distribution make this task endishly complex. Together, these challenges result in a new management complexity that breaks current paradigms based on ad-hoc or centralized and rigid monolithic management applications. Traditional management systems, that are mainly ad-hoc or centralized, reach their limits in such complex management contexts. This PhD work aimes to bring some responses to overcome these di culties by proposing a new management approch based combining advanced software engineering techniques, component based systems, middlewares, and network management systems concepts. Our proposal breaks with current management paradigms mainly related to quite rigid and monolithic systems. To validate our approach, we have developed DASIMA framework (Domain-based Architecture for Scalable Integrated MAnagement middleware). DASIMA middleware framework, implements a new management approach combining domain-based and architecturebased management. It was implemented as a fully component-based management middleware intended to scale-up and down according to the number and the distribution of managed resources. DASIMA has been experimented with a real industrial application in the Machine to Machine (M2M) context. Our experiments were done in reuced context (local network) and large scale deployement context (GRID 5000) and they con rmed the interest of developing adaptable middleware to manage networked systems in multi-scale contexts.