Date of Award
Doctor of Philosophy (PhD)
S. S. Iyengar
The difficulty in debugging distributed programs motivates the development of formal methods for designing distributed programs that are easier to debug and maintain. We address state identification problem for distributed systems using the finite state I/O automaton model. A state S is identified based on the unique event sequences starting at S, called distinguishing sequences. An automaton is diagnosable if every state has a distinguishing sequence. A distributed program may not be diagnosable even if its components are diagnosable. Non-dignosable automata can, in some cases, be converted to a diagnosable form by relabelling some of its transitions in a way that preserves the semantics of the program. Not all automata can be converted to a diagnosable form in this way. This is due to inherent ill-posedness of specification. Two algorithms to convert a non-diagnosable automaton to a diagnosable form are presented. Debugging is the controlled execution of one program by another. The latter is called the supervisor of the former. The supervision operation is defined so that the debugging of a distributed program by distributed debuggers is reduced to the same as the debugging of a single program by a single debugger. An algorithm to construct a debugger for a diagnosable program is developed. Every diagnosable program has a unique debugger associated with it. This leads to the introduction of the notion of debugging complexity of programs.
Nanavati, Amit Anil, "Designing Diagnosable Distributed Programs." (1996). LSU Historical Dissertations and Theses. 6359.