Design and Performance Analysis of a Scalable Backend Architecture for High-Concurrency Web Applications

Abstract

The rapid growth of online services has placed increasing demands on backend systems, particularly in terms of scalability, reliability, and response efficiency. As user traffic continues to rise, traditional monolithic architectures often face performance bottlenecks caused by limited scalability and inefficient resource utilization. This paper proposes a scalable backend architecture that integrates microservices, load balancing, Redis caching, and asynchronous task processing to improve system performance under high-concurrency workloads. The architecture was implemented using Spring Boot, MySQL, Redis, and Nginx, with Docker used for service deployment and management. Performance evaluation was conducted using Apache JMeter under workloads ranging from 500 to 5,000 concurrent users. Experimental results indicate that the proposed architecture maintained an average response time below 240 ms for up to 3,000 concurrent users. At the maximum workload of 5,000 concurrent users, the system achieved a throughput of 2,150 requests per second, representing a 38.6% improvement over the baseline monolithic implementation. The introduction of Redis caching reduced database read operations by approximately 67%, while CPU utilization remained below 72% during peak traffic. In addition, asynchronous processing reduced the execution time of background tasks by 43%, improving the responsiveness of user-facing services. These results demonstrate that the proposed backend architecture can effectively improve scalability and service stability while reducing system resource consumption. The study provides practical guidance for developing reliable backend systems for modern web applications.