Библиографическое описание:Comparative tests of decision making algorithms for a multiversion execution environment of the fault tolerance software : доклад, тезисы доклада / I. Kovalev [et al.]. - Текст : непосредственный // Proceedings - 2017 European Conference on Electrical Engineering and Computer Science, EECS 2017 / 2017 European Conference on Electrical Engineering and Computer Science, EECS 2017 (2017 ; 17.11 - 19.11 ; Bern) : Institute of Electrical and Electronics Engineers Inc., 2018. - P. 211-217. - DOI 10.1109/EECS.2017.47.
Аннотация:many areas are currently experiencing rapid growth, requiring reliable and fault-tolerant software in general and, in particular, management systems. The most effective today is multiversion programming, one of the approaches to improve the reliability of software by introducing redundancy. With this approach, there arises the problem of choosing the right variant from the collection by the answer of multiple versions, which is solved by various decision algorithms, most often based on voting. The reliability of this solution greatly affects the reliability of the redundant system as a whole. The article considers existing decision-making algorithms in multiversion execution environments and suggests own modifications of existing voting algorithms and t / (n-1) algorithm. A software implementation of the simulation environment that implements simulations of versions with specified characteristics and proposed modified algorithms is considered. The results of the simulation are considered, the dependence of the reliability indicators of the system on its input parameters is shown, a comparative analysis of various decision algorithms is made on the basis of simulation results. The obtained results show the effectiveness of the proposed modifications of the voting algorithms by an agreed majority and the specificity of the t / (n-1) decision algorithm, which makes sense to apply not in all systems. Most importantly, the simulation results show the possibility of creating a reliable system from not reliable software components, which is very important, especially for areas where there are currently no highly reliable algorithms or software components.
