17/1998 : Saving Resources in Fault-Tolerant Embedded Real-Time Systems by Using Diverse Software in Probabilistic Time-Redundant Task Execution

RR Number

17/1998

Conference

Workshop on Embedded Fault-Tolerant Systems, May 1998

Author(s)

Abstract

The time-redundant execution of tasks is a mechanism for detecting errors in fault-tolerant systems. Probabilistic time-redundant execution is a CPU-time saving variant of time-redundant task execution that allows a limited percentage of secondary task instances to be aborted before their completion, while guaranteeing that the primary instances always complete.

This paper extends the basic model of probabilistic time-redundant task execution to achieve an even better resource efficiency: It allows different algorithms for the primary and secondary instances of tasks. The presented work provides methods for computing the time consumption of tasks in this model. It presents strategies for selecting algorithms for primary and secondary task instances and illustrates how this model is used to save CPU-time.

Bibtex

@inproceedings{Puschner:EFTS1998,
  author =       "Peter Puschner",
  title =        "Saving Resources in Fault-Tolerant Embedded Real-Time Systems by Using Diverse Software in Probabilistic Time-Redundant Task Execution",
  booktitle =      "Proc. Workshop on Embedded Fault-Tolerant Systems",
  year =         "1998",
  month =        "May",
  pages =      "38-43"
}

Download

Get rr-17-1998.pdf - Adobe PDF-format, (194.74 KB; posted at August 20 2001)