Russian scientists have developed a method to prevent website crashes during a sudden surge of users. For example, when thousands of people simultaneously access a marketplace during a sale or try to take an online exam. Specialists from Togliatti State University and Kabardino-Balkarian State University have created mathematical models that allow them to determine dangerous loads in advance and choose measures to maintain service operability.

Usually, problems begin when the number of requests exceeds the server's capabilities. The website starts to slow down, stops opening, or completely ceases to function. The approach proposed by the scientists allows detecting the approach to a critical limit even before a failure occurs.

To do this, researchers developed a model that takes into account the system's state and provides for an automatic reaction to overload. For example, the service will be able to redirect some requests to the cache, temporarily disable secondary functions, or connect backup servers. According to the authors, such decisions should be made in fractions of a second without human intervention.

The development was tested using mathematical modeling and historical data. Researchers believe that the method can form the basis of systems capable of independently distributing loads and preventing technical problems.

The results were presented at an international conference in Barcelona. Now, specialists plan to create algorithms for mass online testing, where thousands of participants can simultaneously access one resource. Scientists are also ready to cooperate with IT companies for the practical implementation of the technology.