Fine-Grained Implementation of Fault Tolerance Mechanisms with AOP: To What Extent?

The benefits of using aspect oriented programming (AOP) for separation of concerns is well-known and has been demonstrated in many works, including for dependable computing. In this paper, we use this composition capability of AOP to develop micro-aspects that can be combined together to realize a given fault tolerance mechanism. The toolbox of microaspects can be used to make mechanisms easily configurable and by the way to simplify their update. We show that the composition of micro aspects leads to undesirable side effects of the interactions between them, called interferences. We propose an approach to detect interferences with executable assertions, using an extension of AspectJ called AIRIA that enables control over an aspect chain at a shared join point. We finally draw the lessons learnt and discuss to what extent AOP can be used to develop fault tolerance mechanisms.