Objectives. To upgrade and complement the existing load balancing model in multi-server systems, taking into account the hardware characteristics of the server equipment, as well as the most loaded components of the server equipment in the video conferencing service cluster in educational processes and distance education.Methods. The existing mathematical model of load balancing as a mass exchange system is considered, when significant changes are introduced: penalties for equipment downtime and penalties for waiting in a queue will depend on the load on the server hardware components in the cluster architecture of video conferencing service.Results. Formulas are given for calculating the total performance of a cluster of n servers with the maximum and minimum load of server hardware components in a videoconferencing system cluster.Conclusion. A modeling complex has been developed to test the mathematical model on a system of up to n < 10 servers in a cluster of a videoconferencing system. Based on the results of calculations of the modeling complex, it was concluded that it is necessary to upgrate the existing algorithm for balancing the load on the selected BigBlueButton video conferencing service.