MOSCOW, November 7. /TASS/. Researchers from the Siberian Federal University together with their colleagues from Pamukkale University and Yildiz Technical University (Turkey) have created an algorithm that can cut costs on and boost the reliability of IT management systems for organizations, the press office of SFU said.
The study has been published in the journal Software Engineering Trends and Techniques in Intelligent Systems. The new approach to choosing IT management systems for organizations will enable better selection of the best possible components for reliable and financially beneficial multi-version software.
The scientists’ work is devoted to improving those systems controlling industrial processes, for example, the process of manufacturing steel or temperature control in fission reactors. In such cases, the N-version programming (NVP), the software using several versions of one program system to solve the same task, is applied. Given those circumstances, if one version failures, the others continue functioning in a normal way, and a breakdown of the entire system is avoided. The NVP is applied in those cases where security is of critical importance, for example, in space research or in the chemical industry.
An NVP consists of various programs than in turn includes several program modules. Each module has to solve a specific computational task. Traditionally, an algorithm of choosing necessary modules does not take into account any interaction between properties of modules: their prices, reliability, and resource requirements of the software.
SFU programmers suggested a new algorithm for choosing an information management system that incorporates these connections. Such a choice results in a system, which consumes fewer resources but works faster.
The scientists tested the new algorithm on the work of a cargo traffic company from Krasnoyarsk. By applying the algorithm, the freight turnover, depot functioning, and accounts department were optimized. It turns out that the new algorithm has cut overall costs in cases of system breakdown and system downtime (the inactive time of computer facilities, for example, because of a hang) from 4.5 hours to 37 minutes per year.