best-case analysis in real-time system

ECRTS: Exact Best-Case Response Time Analysis of Fixed Priority Scheduled Tasks

motivation

Real-time system analysis has traditionally been focused on the analysis of worst-case behaviour in order to provide guarantees for tasks meeting their deadlines.

There are however situations when the best case behaviour of a task or a system is important as well.

One such situation is in the calculation of response jitter, the maximum variation in the response time of a task or a sequence of tasks.

Typically, response jitter is an important parameter in control systems.

Control theory is dependent on the assumption that sampling and actuation is performed with precise periodicity

and hence large variations in these periods can make an otherwise stable control system become unstable.

In order to guarantee control performance and stability it is therefore important to find tight bounds to the magnitudes of such variations.

A good best-case analysis is a necessity for such tight bounds.

The method is based on local analysis of each node with the response jitter of one task or message being the release jitter of the next task in the sequence.

Since response jitter is the difference between the worst- and best-case response times, finding tight bounds on these response times is important.

While good methods exist for determining worst-case response times of tasks, the best-case has not received the same attention.