Abstract

In real-time systems, one generally identifies two types of timing requirements, hard and soft constraints. In this study, it is assumed that the Hard Real-Time traffic (HRT) is periodic with deadlines that must be guaranteed, while the Soft Real-Time traffic (SRT) is aperiodic with timing constraints that could occasionally be missed without major consequences. In this paper, the problem of scheduling these two types of traffic with different performance objectives will be addressed: (1) ensure that the timing requirements of HRT traffic are met; (2) minimize as much as possible the response time of SRT traffic while satisfying (1). For this purpose, we propose an easily implementable and low-complexity traffic shaping policy, which preserves feasibility and improves response times for SRT traffic. The underlying idea is that it is possible to diminish the response time of SRT traffic if the busy periods induced by the HRT traffic are “harmoniously” distributed over time, creating time intervals during which the resource (i.e. the processor or the medium) can be used by SRT traffic with minimum delay. A computer-implementable algorithm that has to be executed independently on each node of the bus is also provided, as well as several extensions of the original model.

Author Keywords: Real-time systems; Traffic shaping; Scheduling algorithm; Local area network

Article Outline

1. Introduction

1.1. Limitations of existing solutions

1.2. Goal of this paper

1.3. Organization of the paper.

2. Description of the problem

2.1. Framework

2.2. Formulation of the problem

2.3. Minimizing the response time of the aperiodic traffic

3. Traffic shaping with an on-line procedure

3.1. Density and emission sequences

3.2. Deadline first allocation

3.3. Algorithm

4. Generalizations

4.1. When deadlines are smaller than the periods

4.2. Desynchronized stations with known offsets

4.3. Desynchronized stations with unknown offsets

4.4. Jitter in emission

4.5. Variable characteristics of HRT messages: mode changes

4.6. Probabilistic guarantees with unreliable medium

5. Experiments

5.1. Synchronized case

5.1.1. Average response time for SRT traffic

5.1.2. Variance in response times for SRT traffic

5.2. Desynchronized case

5.2.1. Average response time for SRT traffic

5.2.2. Variance in response times for SRT traffic

6. Conclusions

Appendix. Worst-case response time on a priority bus

References