Distributed, Agent-Based fault recognition
and prediction in networked domains
Researcher:
Keith Falconer
Date
completed:
Current, MOD funded.
Papers
published:
None yet.
Abstract:
Networked domains are now complex structures, and
many faults can occur, which can cause them to operate
outside their required quality of service. The aim
of this research programme is to develop a distributed,
agent-based system which can monitor conditions
within complex networking domain, and identify sources
of faults that can occur. This work will also be
extended so that the agent-based system can predict
where faults will occur in the future, and try to
overcome them.
The research will involve several major components:
• Research into fault detection within networked
domains. This will involve researching into networking
faults, such as router and service problems, which
occur with domains, and defining their likely causes.
• Domain model. This will involve the creation
of a model of domain, which identifies the infrastructure
of all the resources, including services, networked
and local devices, and so on. This model will be
used by the agent-based system to create a map of
the complete domain. From this agents should be
able to generate their own tests, and diagnose the
domain.
• Agent model. Each agent will monitor local
environments, and also sense conditions within a
domain (as this allows for a distributed approach
to fault-finding). The elements which are monitored
can range from simple monitoring, such as local
CPU loading, to complex service provision monitoring,
such as e-mail or WWW provision. The monitoring
of the conditions will map to the faults which can
occur within the domain, and will make up the basic
foundation of the system.
• Analytical results. This will involve practical
monitoring a large network for its performance,
and logging when faults occur. These tests will
highlight faults. Along with this IT Professional
will be interviewed to determine the major sources
of faults within large, and complex domains.
• Research model development. This will involve
setting up the fault monitoring and prediction system
to run over a practical domain, and monitor if the
system can accurately prediction, and diagnose faults.
