Today, the number of Internet users reaches more than 50% of all the inhabitants of the Earth (4.5 billion people). It is the most popular tool in the world for obtaining information, communication, communication, advertising and direct sale of goods and services, electronic payments and bank account management, and for many other areas. Hardware functioning of such a complex and multitasking network structure is provided by billions of computers, data centers and servers. Fiber-optic cables, electrical wires, space satellites, as well as infrastructure for wireless information processing are used to connect to the Internet.
The further development of the Internet faces a number of fundamental problems, one of which is the emergence of super–dense networks: an explosive growth in the number of subscribers and the number of their devices and gadgets (phones, tablets, smart watches and household appliances), which consume network information exchange traffic in their work. At the level of local networks, this process threatens to overload the equipment, and private users may face an increase in the time to download information from the Internet.
Moreover, similar situations may occur at large warehouse complexes, as well as industrial facilities and government critical infrastructure that use systems operating through the network. In connection with this problem, the urgency of creating systems capable of quickly redistributing traffic is increasing, creating more efficient schemes for using information infrastructure resources.
"We have developed a mathematical model that allows us to describe the use of traffic at different levels of the Internet – the end user or
local networks – and to predict the increase in network congestion with high accuracy in advance in order to take proactive measures to prevent system overloads," says Oleg Markelov, a leading researcher at SPbSETU "LETI".
During the experiments, scientists collected statistical anonymized information about the behavior of users on the Internet from several sources: from the local network of LETI University, from open data on the traffic of the reference network connecting several dozen universities in Japan, traffic during the 1998 FIFA World Cup was also analyzed.
To find correlations, the researchers studied four main characteristics of the data: where the information was sent from, the destination, the time of sending and the amount of data. Based on the collected statistics, a mathematical model of the relationship of traffic behavior at different levels of the network was developed.
"If earlier the amount of information transmitted through the local network was determined by a simple sum of user activity, then our analysis showed that their behavior can affect traffic on each network node. Using the proposed approach, in the future it is possible to create dynamic routing flow control algorithms that will allow users to avoid long waits for information on the Internet, as well as reduce the load on network equipment and thereby increase its reliability and security," Oleg Markelov notes.
The study was conducted within the framework of the "Priority 2030" program.