Abstract-Autonomic network management is an innovative vision promising new horizons of efficient networking systems free from human control. This promise has, thus far, ushered in enormous yet dispersed research contributions in both industry and academia. The work presented in this article aims at putting these efforts into perspective deriving a more holistic view of the literature in this area. We analyze the requirements and the main contributions for the building blocks of any autonomic network management system (ANMS). We then describe a coherent classification methodology to compare existing ANMS architectures. Based on this analysis, we suggest a reference framework and highlight some open challenges and describe new research opportunities.