A Unified Model for Performance and Reliability of Fault-Tolerant/Multi-Mode Systems.
DUKE UNIV DURHAM NC DEPT OF COMPUTER SCIENCE
Pagination or Media Count:
This paper unifies different models and relates different performance and reliability measures that have been proposed for the analysis of fault-tolerant computer systems. We model the changes in the structure of the system due to different events such as degradation, failure or repair as a continuous time Markov chain. In particular we consider the execution of a job on such a computer system where a service rate or a reward rate is associated with each structure-state. We allow different types of service-preemption interactions due to changes in the structure-state of the system. We derive the distribution of the completion time of a given job. Although the developed technique are suitable for the analysis of complex systems, we demonstrate their use through a simple switching server example.
- Statistics and Probability
- Computer Programming and Software
- Computer Hardware