During the design and maintenance of complex IT systems, the possible failures and unexpected events has to be addressed in order to maintain functionality. With the growing significance and decreasing costs of deploying applications on Cloud infrastructures, the ability to cope with uncertain events, such as component failures or “noisy neighbors”, is getting more important aspect for applications and even critical when deploying high availability applications.
Current thesis introduces a novel method of modelling IT systems by identifying core mechanisms, and exploiting simulation results to cope with imprecise information. Widely used redundancy patterns are introduced to the system model, to describe the potential results of a dynamic reconfiguration from a singular system state.
Visual data analysis techniques are then applied to acquire detailed and intuitive information about the predicted qualitative behavior of the system and to allow the system designer or the operator to preempt system failures by means of dynamic reconfiguration.