PROCESSING MASS USER REQUESTS IN DISTRIBUTED ENTERPRISE SYSTEMS

The purpose of the study is to experimentally determine an architectural solution for processing massive user requests in distributed enterprise systems to improve performance, reduce delays in responding to requests, and ensure scalability.The subject of the study is methods of balancing the load on servers in distributed systems.The article presents an analysis of existing approaches to loading balancing, the formulation and results of computational experiments on load testing of the load balancers according to the defined performance and efficiency criteria.Conclusions: 1) the expediency of using HAProxy and NGINX, as well as the advantages and disadvantages of each, is substantiated; 2) an architectural solution for load balancing in distributed corporate systems based on HAProxy and NGINX is developed, which provides optimality in terms of: performance, scalability, fault tolerance and request processing speed; 3) a test environment for experimental comparison of NGINX and HAProxy is developed.To deploy the system, we used a microservice architecture based on Flask, containerization with Docker, and AWS infrastructure services for cloud deployment.Monitoring and analysis of the load was performed using Prometheus and Grafana, and caching was implemented through Redis; 4) the problem of determining the server for load distribution was solved using machine learning methods.5) experimentally proved the optimality of using HAProxy in terms of performance indicators, in particular, average latency (less by 17.42%), stability of load balancing (less than standard deviation by 77.44%), total number of processed requests (more by 20.39%) and request processing speed (more by 20.43%).Thus, the efficiency of HAProxy is higher in highly loaded systems where the speed and stability of balancing are critical.