Software redundancy: what, where, how

Software systems have become pervasive in everyday life and are the core component of many crucial activities. An inadequate level of reliability may determine the commercial failure of a software product. Still, despite the commitment and the rigorous verification processes employed by developers, software is deployed with faults. To increase the reliability of software systems, researchers have investigated the use of various form of redundancy. Informally, a software system is redundant when it performs the same functionality through the execution of different elements. Redundancy has been extensively exploited in many software engineering techniques, for example for fault-tolerance and reliability engineering, and in self-adaptive and self- healing programs. Despite the many uses, though, there is no formalization or study of software redundancy to support a proper and effective design of software. Our intuition is that a systematic and formal investigation of software redundancy will lead to more, and more effective uses of redundancy. This thesis develops this intuition and proposes a set of ways to characterize qualitatively as well as quantitatively redundancy. We first formalize the intuitive notion of redundancy whereby two code fragments are considered redundant when they perform the same functionality through different executions. On the basis of this abstract and general notion, we then develop a practical method to obtain a measure of software redundancy. We prove the effectiveness of our measure by showing that it distinguishes between shallow differences, where apparently different code fragments reduce to the same underlying code, and deep code differences, where the algorithmic nature of the computations differs. We also demonstrate that our measure is useful for developers, since it is a good predictor of the effectiveness of techniques that exploit redundancy. Besides formalizing the notion of redundancy, we investigate the pervasiveness of redundancy intrinsically found in modern software systems. Intrinsic redundancy is a form of redundancy that occurs as a by-product of modern design and development practices. We have observed that intrinsic redundancy is indeed present in software systems, and that it can be successfully exploited for good purposes. This thesis proposes a technique to automatically identify equivalent method sequences in software systems to help developers assess the presence of intrinsic redundancy. We demonstrate the effectiveness of the technique by showing that it identifies the majority of equivalent method sequences in a system with good precision and performance