Automatic workarounds as failure recoveries

Mechanisms to automatically recover from problems are key elements to designing self-managed software systems. So far most research on self-managed systems focused on non-functional problems, such as architectural mismatches, performance problems and configuration incompatibilities. In our work, we focus on techniques for automatically recovering from functional failures. We aim to exploit the intrinsic redundancy of many complex software systems that can produce the same results in several ways. We are investigating techniques that, in case of failure, look for execution sequences that are equivalent to the one that leads to the failure, and can thus be executed in alternative to the failing one to produce the expected results. We refer to sequences that are equivalent to failing ones and can recover from failures as automatic workarounds. The contribution of the thesis will be the definition of techniques to automatically generate equivalent sequences from different models, and to identify automatic workarounds within sets of equivalent sequences, and a thorough evaluation of the techniques in the context of self-healing software systems.