Related Research
The first steps of the research on non-locking distributed protocols were made about two decades ago [14]. As expected, such a fundamental problem received much further attention (cf. e.g. [9, 10, 17, 19]), part of the e ort being towards classifying system models and commu- nication primitives, according to their non-blocking synchronisation power. According to the classi cation, there exist objects, which, used in so called universal constructions, can implement any other object of any class (cf. e.g. [9, 10]). Universal wait-free and lock-free methods are expensive, so a strong stream of research is towards deriving efficient implementations of speciffic objects (based on the possibility results implied by the universal constructions). Another stream of research e ort involves the introduction of new architectural primitives for lock-free process synchronisation, such as the transactional memory [11], which allows programmers to de ne customised operations to be applied to multiple, independently chosen words of memory.
Actually, the first suggestion for the wait-free approach to real-time communication was rst discussed at least two decades ago [22, 23], but was "lost" in the real-time systems community until recently, when it was revived by Kopetz and Reisinger [13], followed by a series of interesting results by Anderson et.al. (including [2, 3, 4, 5]), and the more recent work by Chen and Burns [6], in applications such as automotive system control, real-time database systems, video-conferencing systems and distributed real-time systems.
<< previous: Main Idea top