Self-Adaptive Model-Based Performance Management in Industrial Data Centers

System Management of large Data Centers addresses both the initial provisioning of software systems on top of the available infrastructure and the operational management of resources at runtime. The system management aims to optimize the provider's defined objectives, while satisfying the given system and customers constraints. In this paper, we propose an autonomous system management approach, which incorporates, if available, software models of business applications and the applications' management experiences from the past. We discuss existing approaches and present an extended management architecture that is built and trained at design-time and calibrated at runtime.