Abstract (2007-02-08)
Real-time systems are often constructed based on a set of real-time
tasks. These tasks may be scheduled and executed according to given
release patterns. There have been a number of methods and tools
developed for timing analysis to estimate the worst case (and also
the best case) response times of computation tasks for such systems,
e.g., Rate-Monotonic Analysis for periodic tasks, Real-Time Calculus
for tasks described using arrival curves, and TIMES using Timed
Automata.
Some of these techniques, e.g., implemented in TIMES can deal with
systems with complex release patterns, but do not scale well with
system size and complexity; the others are scalable but can not handle
systems with complex structures. Our goal is to take the advantages
of these existing techniques and develop an analysis technique, that
is scalable and also capable of handling complex systems. The essential
idea of our approach is to abstract the timed automata nodes using
arrival curves, which can be done modularly for each node, and to
compose the analysis results in order to perform the system level
analysis.