Findings from the First Annual Storage and File Systems Benchmarking Workshop

Single self-adaptive systems can now be built, but systematic methods should be developed for building these systems. Systematic methods require good models for prediction, control, error detection/fault diagnosis, and optimization. Models must describe behavior at different time and detail scales, for different tasks (e.g., energy, error detection) and for different degrees of accuracy. Models can be self-learned or provided by expert human engineers. Models should describe both the system and its environment. Objectives need to be clearly defined for accountability, performance, and reliability of self-adaptive systems.