Stanko P. Methods for optimizing the discipline of service in real-time operating systems

The information and control and calculating network of a large production enterprise with a set of territorially distributed branches is considered. The dissertation is devoted to the creation of hardware and software methods for optimising the service disciplines of real-time operating systems, secure event processing in distributed operating systems of computer networks with different physical transmission media. The computer network as an integral part of a modern robotic enterprise plays a crucial role in managing the production process in general, coordinating the time parameters of its autonomous parts, synchronization of production, logistics and other tasks. Accordingly, the software of the manufacturing enterprise has its own specific features that are not inherent in general purpose software. The methods of analysis and monitoring of a key element of production software - real-time operating system are considered in the work. Particular attention is paid to hard real-time operating systems. According to the results of processing a large amount of statistical data of general-purpose software and production software, one of the key parameters of the production software production is tolerance to memory access delays. It is established that the tolerance to the access delays to the buffer memory of the computing device is a crucial characteristic of the rough real-time systems. To obtain quantitative characteristics of production software with uncontrolled anomalies of software products, a combined statistical and lexical approach to measuring the characteristics of production software has been developed - modified Halsted metrics with embedded stochastics. A method for estimating modified metrics and obtaining effective and unbiased estimates has been developed. In the dissertation work new approaches to OS reliability and security management are developed: creation of monolithic structure of OS kernel (including RF OS); use of the concepts of multi-service and virtual machine; use of special programming languages tolerant to unauthorized intrusions. It is established that the concentration of all efforts only on the optimisation of the structures of the OS kernel (micro kernel or monolithic kernel) does not give a general dissolving to the contradictions between reliability and performance. Only a comprehensive approach to solving the problem will give the desired result. Network anomalies are classified according to the degree of their danger and negative impact on the functioning of computerized and robotic production. Statistical characteristics of anomalies are allocated and methods of their detection and neutralization are developed. The optimal method of dynamic polling of the RF operating system, which processes with different degrees of interruption, has been developed. Their time characteristics are analysed and the conditions of observance of the critical service life are formulated. A method for optimising the discipline of intermittent service discipline with the use of the method of frequency-monotone analysis as the most effective method of analysis and synthesis of optimal service disciplines of the RF OS with priorities has been developed. The set of developed methods makes it possible to improve time characteristics, tolerance to delays, reliability and security of production software and its key element - the real-time operating system.