In this paper we investigate whether the statistical Worst Case Execution Time (WCET) estimation methods devised for embedded platforms can be successfully applied to find the Worst Case Response Time (WCRT) of a network application running on a complex hardware platform such as a contemporary commercial off-the-shelf (COTS) system. Establishing easy-to-use timing validation techniques is crucial for real-time applications and meeting the Quality of Service requirements in general. We study nondeterminism of task execution times and exploit it by the application of redundant computations to achieve better fit of the model and lower WCRT bounds. We use many instances of the network-based RT application, expose them to identical streams of events and study correlation between response times and how it evolves over time. Experiments help us determine if states of separate CPUs will become synchronized which could affect the correlation. We also take a look on self-similarity in the distribution of response times. Test are conducted for many state sizes and with and without additional tasks on the tested system to examine the impact execution environment has on our results
Authors
Additional information
- Category
- Publikacja w czasopiśmie
- Type
- artykuły w czasopismach recenzowanych i innych wydawnictwach ciągłych
- Language
- angielski
- Publication year
- 2014
