Browse Prior Art Database

Reflexive Fault-Tolerance, Knowledge Discovery Technique for Enhancing Fault-Tolerance

IP.com Disclosure Number: IPCOM000117626D
Original Publication Date: 1996-Apr-01
Included in the Prior Art Database: 2005-Mar-31
Document File: 4 page(s) / 80K

Publishing Venue

IBM

Related People

Maitan PhD, J: AUTHOR [+2]

Abstract

Disclosed is a method for enhancing fault-tolerance in certain control systems. The method uses knowledge discovery techniques to derive an algorithm based on dependencies in the set of input data sources. The algorithm provides an alternate source of input data for a failed input based on the dependency relationship that exists between the (failed) input's characteristics and the characteristics of the inputs that continue to operate correctly.

This text was extracted from an ASCII text file.
This is the abbreviated version, containing approximately 52% of the total text.

Reflexive Fault-Tolerance, Knowledge Discovery Technique for Enhancing
Fault-Tolerance

      Disclosed is a method for enhancing fault-tolerance in certain
control systems.  The method uses knowledge discovery techniques to
derive an algorithm based on dependencies in the set of input data
sources.  The algorithm provides an alternate source of input data
for a failed input based on the dependency relationship that exists
between the (failed) input's characteristics and the characteristics
of the inputs that continue to operate correctly.

      Disclosed is a technique that provides a method of enhancing
fault-tolerance in certain control systems.  This technique uses
knowledge discovery techniques to analyze system input data and find
dependencies that will allow algorithms to be constructed such that
an algorithm can replace a failed device as the source of system
input.  The acceptability of this technique in any given application
is dependent upon the degree of the dependency relationship; total
dependencies will provide seamless replacement of the failed input
source, partial dependencies will result in some degradation.

      Fig. 1 shows a simplified control system that uses three inputs
to calculate the system output.  To implement reflexive
fault-tolerance in such a system, data from inputs X, Y, and Z would
be collected at regular intervals, time-stamped and stored.  The data
would be processed using a knowledge discovery technique (such as
rough set theory's decision logic)(1) to uncover dependencies that
may exist among the input sources.  Assume that data from input Z
exhibits dependency relationship to a combination of characteristics
of data from input X and input Y.  Based on this dependency
relationship, an algorithmic model is generated that can provide an
alternate source (derived) for input Z, in the event that input Z
fails.  Fig. 2 shows this new implementation providing the additional
fault-tolerance that has now been implemented for input Z.  It is
possible to implement these techniques...