2026-04-13T01:47:28Zhttps://recercat.cat/oai/request

oai:recercat.cat:2117/3310122025-07-16T22:26:07Zcom_2072_1033col_2072_452949

Modelling and predicting extreme behavior in critical real-time systems with advanced statistics Vilardell Moreno, Sergi Serra Mochales, Isabel Cazorla Almeida, Francisco Javier Àrees temàtiques de la UPC::Informàtica::Arquitectura de computadors High performance computing CRTES MBPTA Extreme Value Theory Càlcul intensiu (Informàtica) In the last decade, the market for Critical Real-Time Embedded Systems (CRTES) has increased significantly. According to Global Markets Insight [1], the embedded systems market will reach a total size of US $258 billion in 2023 at an average annual growth rate of 5.6%. Their extensive use in domains such as automotive, aerospace and avionics industry demands ever increasing performance requirements [2]. To satisfy those requirements the CRTES industry has implemented more complex processors, a higher number of memory modules, and accelerators units. Thus the demanding performance requirements have led to a merge of CRTES with High Performance systems. All of these industries work within the framework of CRTES, which puts several restrictions in their design and implementation. Real Time systems require to deliver a response to an event in a restricted time frame or deadline. Real-time systems where missing a deadline provokes a total system failure (hard real-time systems) need satisfy certain guidelines and standards to show that they comply with test for functional and timing behaviour. These standards change depending on the industry, for instance the automotive industry follows ISO 26262 [3] and the aerospace industry follows DO-178C [4]. Researches have developed techniques to analyse the timing correctness in a CRTES. Here, we will expose how they perform on the estimation of the Worst-Case Execution Time (WCET). The WCET is the maximum time that a particular software takes to execute. Estimating its value is crucial from a timing analysis point of view. However there is still not a generalised precise and safe method to produce estimates of WCET [5]. In the CRTES the estimations of the WCET cannot be lower than the true WCET, as they are deemed unsafe; but they cannot exceed it by a significant margin, as they will be deemed pessimistic and impractical. 2020-05 Conference report Vilardell Moreno, S.; Serra Mochales, I.; Cazorla Almeida, F.J. Modelling and predicting extreme behavior in critical real-time systems with advanced statistics. A: . Barcelona Supercomputing Center, 2020, p. 37-38. https://hdl.handle.net/2117/331012 en eng Open Access Attribution-NonCommercial-NoDerivs 3.0 Spain 2 p. application/pdf application/rdf+xml; charset=utf-8 Barcelona Supercomputing Center