A process architecture and runtime environment for dependable distributed applications

Differences among applications preclude the development of a single fault tolerance mechanism that is suitable to all possible types of applications. These differences can be found in, among other areas, application functionality and dependability requirements. This research focuses on the design and implementation of a flexible software-implemented fault tolerance (SIFT) environment constructed from ARMOR (Adaptive Reconfigurable Mobile Objects of Reliability). The ARMOR-based SIFT environment hosts specific fault tolerance mechanisms that provide error detection and recovery services to user applications. New fault tolerance mechanisms can be designed as element modules that plug into the reliable ARMOR runtime environment. An ARMOR process is designed to be generally reconfigurable, meaning that virtually any aspect of its functionality—including the fault tolerance mechanisms that protect the user applications—can be changed by changing the elements that make up the ARMOR process.; This thesis introduces the theory behind (1) the construction of ARMOR processes and (2) the behavior of ARMOR processes during runtime. Reconfigurability properties of the ARMOR architecture are examined, including the ability to establish criteria that must be satisfied for a proposed reconfiguration to be considered “safe” with respect to the current ARMOR's configuration. Next, this thesis describes the SIFT environment constructed from ARMOR processes, including an in-depth case study in which the SIFT environment is used to protect spaceborne, scientific applications for the Jet Propulsion Laboratory (JPL). Error injection experiments validate the SIFT environment's ability to recover not only from user application errors, but also ARMOR errors as well. Finally, five other case studies illustrate how the ARMOR-based SIFT environment provides fault tolerance to user applications from several different domains, each requiring a different set of fault tolerance mechanisms to achieve its desired level of dependability.