Search

The purpose of this report is to outline the major concepts and developments in the area of fault tolerant computing. Both hardware and software fault tolerance issues are addressed. The topics covered include module function and system-level fault detection methods, redundancy and reconfiguration strategies, valid fault models, and coding and checking in computer systems. Software fault tolerance methods such as recovery blocks, design diversity, and checkpointing and recovery are also discussed. Major issues in modeling and evaluation of fault-tolerant systems are outlined. The design of two successful commercial systems is discussed.

This paper describes a measurement-based performability model based on error and resource usage data collected on a multi-processor system. A method for identifying the model structure is introduced and resulting model is validated against real data. Model development from the collection of raw data to the estimation of the expected reward is described. Both normal and error behavior of the system are characterized. The measured data show that the holding times in key operational and error states are not simple exponentials and that a semi-Markov process is necessary to model the system behavior. A reward function, based on the service rate and error rate in each state, is then defined in order to estimate the performability of the system and to depict the cost of different types of errors. Keywords: Computer modeling, Computer systems, Risk assessment, Predictability, Errors, Workload, Measurements, Semi-Markov, Performability, Work measurement., Sponsored in part by Grant NAG-1-613.

This paper describes an analysis of hardware related software errors on the MVS operating system at the Center for Information Technology at Stanford University. The study first examines the software error detection mechanism with particular reference to the detection of software errors related to temporary and permanent hardware problems. About 11 percent of all software errors and over 40 percent of all software failures were found to be hardware related. It is shown that the system is seldom able to diagnose the fact that a software error may be hardware related. Key patterns in the occurrence of hardware related software errors are determined and their effect on system recovery examined. In a HW/SW record, both the hardware and the software errors occur in large clusters and have a significant percentage of lost records associated with them. The system recovery management is less likely to recover from hardware related software errors than software errors in general. It is suggested that the error patterns found in this study could form the basis for the detection and recovery management of hardware related software errors. (Author), Also available as Rept. no. CSL-TN-83-231.