The OASIS Kernel: A Framework for High Dependability Real-Time Systems

Abstract

This paper presents the design and some aspects of implementation of a highly dependable, safety-oriented kernel for real-time applications. It is specifically designed as an execution facility for a deterministic semi-formal model -the OASIS model-which allows to express and verify temporal behaviors and communications of a safety critical real-time application.This paper shows specifically how, from a formalism, and a Domain Specific Language, we achieved to build a generic execution layer that conforms to the highest levels of safety, how the safety is implemented thank to the interaction between the kernel and the compilation tools, and how performance was optimized within these constraints.

Hippo: A formal-model execution engine to control and verify critical real-time systems

“…A similar motivation can also be found in the design choices behind OASIS [Louise et al, 2011], a framework provided by the CEA LIST to generate an executable based on a time-triggered approach. The temporal information in an OASIS model is directly specified in the code using a dedicated language, called ψC.…”

Static and Verifiable Memory Partitioning for Safety-Critical Systems

“…The closest work to ours seems to be described in papers from Camier et al [8] and Louise et al [9], who mention that spatial partitioning of their real-time kernel OASIS is performed by loading MMU configuration tables that are statically generated at compile-time. However, the generation process is not explained.…”

Correct-by-Construction Parallelization of Hard Real-Time Avionics Applications on Off-the-Shelf Predictable Hardware

“…A few approaches have gone further by letting timing analysis results guide mapping and code generation under simplifying hypotheses common in real-time scheduling, e.g., assuming that task WCET values include overheads related to parallel/concurrent execution. Among these approaches, we cite the industrial tool Asterios Developer from KronoSafe, based on the ΨC language [37], as well as the academic tools and toolboxes SynDEx [55], BIP [7], SchedMCore [52], Prelude [42], Lopht [14,18], SigmaC [3,10], the work on the time-triggered mapping of Lustre [15], Xoncrete [11], or the work of Baruah et al on the synthesis of multi-core cyclic executives [23]. We defer the reader to Ref.…”

“…6 The respect of execution time bounds for normal conditions can then be achieved through tight control of scheduling, memory allocation, and synchronization [12]. This is different from time-triggered approaches [15,20,37], which place timing predictability first, potentially at the expense of functional determinism and robustness.…”