Mobile Cloud Computing paradigm has arisen as a major proposal to address collaboration support in working environments. Particularly, this paradigm has proven to be useful in emergency scenarios, education or tourism. However, these environments are commonly based on dynamic network topologies, which imply unstable connections (disconnections and network partitions). In consequence, the availability of the services can be compromised. Therefore, approaches based on the Service Oriented Architecture (SOA) are insufficient and they must be complemented with techniques and methods of Autonomic Computing, in order to ensure the quality attributes of the system against context changes. In this work, a self-adaptive architecture is proposed in order to address the availability of the services deployed in dynamic network environments. This architecture has been designed to provide a common basis for collaborative mobile systems. The architecture follows a component-based design, and it provides a distributed approach to support the dynamic replication and deployment of services. Further, an example scenario based on a real Mobile Forensic Workspace is described to show the applicability of the proposal.