Toward Sofware Synthesis for Distr ibuted Applicat ions

This paper describes Sage, a software environment supporting software development, synthesis, and testing for distributed computing applications. While the principal domain of interest is applications that must be fault-tolerant (i.e., be able to withstand the failure of some of the participants) Sage is not limited to this; it can be extended to distributed applications with no criticality requirements and to those with security requirements. Sage mechanically applies specialized knowledge-theoretic analyses to a distributed application's high-level specification to automatically derive the necessary communication between the participants in the computation. In particular, Sage implements the results of Chandy and Misra [2] and mimics the analyses of others [9, 10, 12, 15] which have previously only been performed theoretically. Sage applies these results to strategies commonly used by programmers of distributed applications, and commonly provided by packaged subsystems for distributed computing (also called "middleware"), to derive and synthesize correct, efficient solutions. Using the modal logic of knowledge to describe and reason about coordinating distributed entities is well accepted [2, 3, 5, 6, 9, 10, 12, 15], if highly specialized, and the semantic model has been adapted to suit different distributed environments [6, 7, 10, 11]. In particular, epistemic logic has been used to optimize solutions, to prove impossibility, and to prove possibility [2, 5, 7, 9, 10, 12, 13]. Despite the range of applications in the literature, no software development environment exists that applies and implements these techniques to facilitate writing distributed applications.