ECONOMIC AND MATHEMATICAL MODELING OF THE LOAD INCREASE OF HIGH-LOAD SYSTEMS

Анотація

An analytical and forecasting simulation of the main indicator of the information infrastructure load measurement - the number of requests to the system per second was carried out in the article. The statistics of this indicator are investigated in order to find and identify the trend of movement of its time series for ensuring the reliability and accessibility of services of IT-company. The relevance of this research is getting some resonance in the information technology industry at this time, because expanding information systems to the ability to transmit petabytes of data per second requires a clear plan for their infrastructure. Providing a minimal forecast and data estimate can not only guarantee the stability of a software product or technology, but also directly affect the cost of production and the amount of profit received.  The timely detection of system crashes can prevent the domino principle, which can cause harm to other synergistic processes of the company. The long-term outlook provides companies with some retreat paths to scale their infrastructure size and provides some time to make decisions about purchasing equipment or expanding their capabilities at “Cloud Services”.

To this end, the Holt forecast was constructed on the basis of statistics data, the statistical detection of abnormal emissions and their removal by the moving average method were performed. Data were grouped to increase the forecast periods and the possibility of using the Holt method was tested. The residuals were analyzed, which showed the adequacy of the model. The results of the forecast are presented in the form of generated reports of the program mathematic product. In the final stage, the results covered 16 forecast periods, which are generally equivalent to two full days. Estimates of cost savings for the average physical configuration of the receiving party and the minimum client software have been announced. On the basis of these data, it was concluded that the importance of conducting research data using such methods of forecasting and analysis of data for different vectors of activity of an IT company.