By separating the process logic from the application code process management systems (PMS) offer promising perspectives for automation and management of business processes. However, the added value of PMS strongly depends on their ability to support business process changes which can affect the process type as well as the process instance level. This does not only impose challenging conceptual issues (e.g., correctness of process schemata after changes) but also requires sophisticated implementation concepts with respect to efficient algorithms, flexible architectures, and reasonable treatment of resources. In this paper we sketch the general implementation concepts for representing process type and process instance data as well as for realizating process schema evolution. All these concepts have been developed and are currently implemented in the ADEPT2 prototype within the AristaFlow project.