A software-implemented fault injection methodology for design and validation of system fault tolerance
Authors:
- W.S. KimWKW.S. Kim
- G. KhanoyanGKG. Khanoyan
- L. CallumLCL. Callum
- A. AgrawalAAA. Agrawal
- J.J. BeahanJBJ.J. Beahan
- 13 November 2002
- conference paper
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 501-506
Abstract
Presents our experience in developing a methodology and tool at the Jet Propulsion Laboratory (JPL) for software-implemented fault injection (SWIFI) into a parallel-processing supercomputer which is being designed for use in next-generation space exploration missions. The fault injector uses software-based strategies to emulate the effects of radiation-induced transients occurring in the system hardware components. JPL's SWIFI tool set, which is called JIFI (JPL's Implementation of a Fault Injector), is being used in conjunction with an appropriate system fault model to evaluate candidate hardware and software fault tolerance architectures, to determine the sensitivity of applications to faults, and to measure the effectiveness of fault detection, isolation and recovery strategies. JIFI has been validated to inject faults into user-specified CPU registers and memory regions with a uniform random distribution in location and time. Together with verifiers, classifiers and run scripts, JIFI enables massive fault injection campaigns and statistical data analysis.Keywords
This publication has 4 references indexed in Scilit:
- REE: a COTS-based fault tolerant parallel processing supercomputer for spacecraft onboard scientific data analysis2003
- Detailed radiation fault modeling of the Remote Exploration and Experimentation (REE) first generation testbed architecture2002
- NFTAPE: a framework for assessing dependability in distributed systems with lightweight fault injectors2002
- Xception: a technique for the experimental evaluation of dependability in modern computersIEEE Transactions on Software Engineering, 1998