FORTS: Fault Tolerance in Real-Time systems through
Scheduling
FORTS:
Fault Tolerance in Real-Time Systems
Through Scheduling
This is a joint project between
Honeywell Technology Center (HTC)
and
Department of Computer Science,
University of Pittsburgh, and is funded by
DARPA / ITO Embeddable Systems Program.
The Investigators of the project at HTC are:
- Sunondo Ghosh (sghosh@htc.honeywell.com)
- Walt Heimerdinger (walt@htc.honeywell.com)
The Principal Investigators of the project are:
- Rami Melhem,
University of Pittsburgh (melhem@cs.pitt.edu)
- Daniel Mosse',
University of Pittsburgh (mosse@cs.pitt.edu)
- Neeraj Suri,
New Jersey Institute of Technology (suri@cis.njit.edu)
Introduction
The University of Pittsburgh (Pitt) is conducting research on scheduling
techniques for real-time systems to ensure that task deadlines continue to
be met in the presence of faults. The algorithms under study offer the
opportunity to increase the reliability/availability of real-time
applications without incurring the overhead of large-scale hardware
redundancy.
The Honeywell Technology Center (HTC) will support the research effort at Pitt
by identifying real-time applications that can be expected to benefit from the
algorithms. This will be based on HTC's experience both with real-time
applications produced by the various Honeywell operating divisions and with
real-time scheduling technology.
Tasks for the subcontract
The following tasks will be performed as a part of the subcontract:
Several existing real-time applications will be considered for applicability
of the fault tolerant scheduling techniques developed at Pitt. This
information will be used as input for the development of modified scheduling
algorithms at Pitt.
Task 2 - Feedback to Pitt
For each application studied as a part of Task 1, the assumptions made by Pitt
research will be considered. The assumptions that need to be relaxed to adapt
the research for realistic applications will be listed. The reasons for
requiring the modification to the assumptions will also be stated.
Task 3 - Evaluate ongoing research at Pitt
Honeywell will evaluate the potential for incorporating new algorithms
developed at the University of Pittsburgh and their corresponding assumptions
in existing and future versions of Honeywell products and applications.
Task 4 - Define a Demonstration
One application will be studied in greater depth and a demonstration showing
the benefits of using the approaches developed at Pitt for the selected
application will be specified. Honeywell will specify a task set to be used
for this demonstration that provides a realistic, but minimal subset of the
typical functions of the application class.
Task 5 - Implement Demonstration Task Software Modules
Honeywell will implement a set of application software modules for the
demonstration defined in Task 4. These application modules will be
representative of the specified application with respect to input/output
behavior and scheduling constraints. They will be designed to be compatible
with the fault recovery algorithms developed at Pitt.
Contact Information
Sunondo Ghosh
Honeywell Technology Center,
3660 Technology Drive, MN65-2600
Minneapolis, MN 55418-1006
Email: sghosh@htc.honeywell.com
Phone: (612) 951-7513
Fax: (612) 951-7438