Presentation title

Probabilistic real-time: restart from scratch

Authors

Federico Reghenzani and William Fornaciari

Institution(s)

Politecnico di Milano, Italy

Presentation type

Technical presentation

Abstract

Obtaining a safe but tight Worst-Case Execution Time (WCET) has become a challenging problem for modern architectures. The growing computational power demand of embedded systems, in addition but opposed to the reaching of technology limits, is increasing the hardware complexity of processors – such as the introduction of many-cores, multi-level caches, complex pipelines, hardware managed interrupts, etc. – that consequently leads to hindering the use of classical WCET estimation techniques. The problem is even magnified when dealing with Commercial-Off-The-Shelf (COTS) components, mixed-criticality and complex operating systems. Probabilistic (hard) real-time has been proposed as a solution to WCET estimation in the previously described scenarios. The measurement-based approach is founded on the well-known Extreme Value Theory (EVT), typically applied to natural disaster prediction use cases, e.g. to estimate the probability of unseen catastrophic floods. The research works in the last years tried to propose hardware implementations compliant with the EVT hypotheses, however, several theoretical challenges are still open and represent the major problem to obtain scientific and industrial consensus.

---

Additional material

• Presentation slides: [pdf]

For more details on this presentation please click the button below: