Abstract
Cloud computing is an emergent technology in the process of becoming ubiquitous. This requires strategies to deal with challenging situations. While a P2P structure is suitable under usual circumstances, other structures may be required in case of strong perturbations and disruptions. This paper describes a partial blackout scenario: the Cloud uses self-management properties to switch from a peer-to-peer structure to a temporary centralised structure, and then returns to normal. The system remains adaptive at all times, while maintaining performance under aggravated conditions. To achieve such self-management, a specific design pattern is suggested. Properties for self-adaptive and self-managing system are described, and implementation perspectives are discussed.