On the use of COTS microkernels for dependable systems

This paper addresses the problem of using COTS microkernels for the design and implementation of dependable systems. Although not designed to deal with faulty situation, their analysis in the presence of faults is of high interest since all upper layers rely on their correct behavior. Such an information can be used to design upper layers accordingly. However, as shown in many works, the results obtained using fault injection on COTS executives raise unacceptable situations for critical applications. The definition of fault containment mechanisms is mandatory to deal with such situations and make simpler the design of upper layer software. The approach proposed in this paper relies on the modeling of microkernel functionalities. Modeling is indeed possible because these functions are often simple. The implementation of efficient fault containment wrappers proposed here is based on a novel approach taking advantage of the notion of reflective components.