An Examination of Two Fault-Tolerant Architectures,
Abstract
Two fault-tolerant computer designs were examined. For each design a functional simulator was implemented and an executive program, recovery software, and application program was coded. The major concern in the executive program development was the handling of input/output and interrupts in the presence of faults. Similarly, the development of the recovery software revealed that the preservation of the application was more difficult than the recovery of the hardware itself. The application was selected from the Titan 3C flight program. A simple compiler was developed to generate the application program code and automatically insert rollback points. This approach eliminated any concern for fault tolerance on the part of the application programmer. However, a significant overhead in terms of memory space and execution time due to fault tolerance resulted. Each design was examined at a functional level relative to its computation capabilities and effectiveness in providing fault tolerance. Weak points were identified in each design and recommendations for correcting them were provided. Neither design completely handled catastrophic failures. (Author)
Document Details
- Document Type
- Technical Report
- Publication Date
- Aug 01, 1973
- Accession Number
- AD0766517
Entities
People
- David K. Switzer
- Frank J. O'brien
- Joseph A. Lauro