A methodology for constructing a stabilizing crash-tolerant application

This paper is an exercise to construct a stabilizing mutual-exclusion protocol that withstands a single crash-failure. We begin with a collection of distributed processes arranged in a ring. The resulting protocol is stabilized by construction. Stabilizing protocols converge to a correct behavior regardless of their initial state. A faulty process is automatically removed from the system and, after repair, automatically integrated into the system. Our technique can be generalized to different systems by substituting appropriate protocols for various components.<
>