Most current software management solutions are missing a systematic and holistic approach for global system management, and apply only to specific system components. This limits the system k extensibility, and does not meet interoperability requirements from the growing heterogeneity of the operating environment. This paper discusses an alternative approach to common management of evolving distributed software, impired by the research in the field of open systems and cybernetic models. The approach involves adapting Stanford Beer's Yiable System Model (VSM) to fhe concrete needs of distributed sofware, supporting evolutionary integration of new functionality, while preserving system stability. We introduce meta-data bindings to existing information models as a basis for management of hierarchical and recursive sofware elements, and an architecture for composition of interacting components with verification of their capabilities during deployment and runtime. The architecture is designed to satisfy criteria of the Eable System Model and to assure conditions necessary for autonomic behavior.