Possible risks with EVT-based timing analysis: an experimental study on a multi-core platform

Abstract

Hardware architectures equipped with multiple cores, cache memory and branch prediction usually preclude the application of classical methods for determining execution time bounds for real-time tasks. As such bounds are fundamental in the designing of real-time system, Measurement-Based Probabilistic Timing Analysis (MBPTA) has been employed. A common choice is towards the derivation of probabilistic Worst-Case Execution Time (pWCET) via the use of Extreme Value Theory (EVT), a branch of statistics for modeling the maximum of a random variable. However, pWCET estimations are usually reported taking a controlled or simulated environment. In this paper we rather apply MBPTA in a real multi-core platform, namely Raspberry Pi 3B, taking into consideration possible interference due to operating system and concurrent activities. The found results indicate that although EVT is a robust technique, it does not always produce adequate models and coherent pWCET estimations. As MBPTA is primarily called for when classical methods are not applicable, as it is the case for the studied platform, the results reported in this paper highlight possible risks in when applying MBPTA for pWCET estimations.