Improved response-time analysis calculations

Abstract

Schedulability analysis of fixed priority preemptive scheduled systems can be performed by calculating the worst-case response-time of the involved processes. The system is deemed schedulable if the calculated responsetime for each process is less than its corresponding deadline. It is desirable that the Response-Time Analysis (RTA) can be efficiently performed. This is particularly important in dynamic real-time systems when a fast response is needed to decide whether a new job can be accommodated, or when the RTA is extensively applied, e.g., when used to guide the heuristics in a higher level optimiser. This paper presents a set of methods to improve the efficiency of RTA calculations. The methods are proved correct, in the sense that they give the same results as traditional (non-improved) RTA. We also present an evaluation of the improvements, by applying them to the particularly time-consuming traffic model used in RTA for ATM communication networks. Our evaluation shows that the proposed methods can give an order of magnitude reduction of the execution time of RTA.

Exact quantification of the sub-optimality of uniprocessor fixed priority pre-emptive scheduling

“…The length of the busy period i w , can be computed using the following fixed point iteration (Audsley et al, 1993), with the summation term giving the interference due to the set of higher priority tasks hp(i). w are monotonically nondecreasing with respect to the iteration count m (Tindell, 1994), and so the fixed point iteration is guaranteed to converge to the worst-case response time i R , provided that the overall taskset utilisation is less than or equal to 1, and the initial value 0 i w is a lower bound on i R (Sjodin and Hansson, 1998).…”

Maintaining the Feasibility of Hard Real–Time Systems with a Reduced Number of Priority Levels

“…(2). Numerous variants of RM schedulability constraints were proposed by Bini and Buttazzo (2004), Bini et al (2008), Min-Allah et al (2007), Lehoczky et al (1989), Tindell et al (1994), Audsley et al (1993), Sjodin and Hansson (1998), and Davis et al (2008) for determining RM feasibility analysis that generally falls into major classes: scheduling points tests Bini et al, 2008;Min-Allah et al, 2007) and response time based tests (Audsley et al, 1993;Joseph and Pandya, 1986;Davis et al, 2008). The latter are superior over the former from analysis time perspectives, because in scheduling points tests feasibility is tested at all scheduling points for all tasks in the set, while iterative techniques have the advantage of making larger jumps in t that result in skipping a large number of scheduling points and hence the feasibility of a task is determined much early; however, scheduling point tests are considered to be the fundamental ones and can be used at the system design stage (Bini et al, 2008), and hence are the focus of this work.…”

“…Let us also denote by J i the first job generated by the task τ i . It is assumed in this work that each task must generate at least one job in a given time slot, and J i has the maximal response time compared with all other jobs generated by τ i Tindell et al, 1994;Sjodin and Hansson, 1998) 1 .…”

Efficient Exact Schedulability Tests for Fixed Priority Real-Time Systems

“…In 1998 Sjodin and Hansson [13] extended Audsley's work, by accounting for blocking factors in the initial value calculation. They also introduced a closed form lower bound on the response time that could be used as an effective initial value.…”