Optimizing Backend Service Performance Through Multi-Level Caching and Database Query Optimization

Abstract

As the number of concurrent users continues to increase, backend systems often experience performance degradation caused by frequent database access and inefficient query execution. This paper investigates the impact of multi-level caching and database optimization techniques on the performance of backend services. The proposed architecture combines Redis caching, SQL query optimization, connection pooling, and asynchronous request processing to reduce database workload and improve response efficiency. A prototype system was developed using Spring Boot, MySQL, and Redis, and evaluated under different workload conditions using Apache JMeter.

The experiments were conducted with workloads ranging from 200 to 3,000 concurrent users. Before optimization, the average API response time reached 486 ms at 3,000 concurrent requests, while the throughput was approximately 1,280 requests per second (RPS). After applying the proposed optimization strategy, the average response time decreased to 214 ms, and the system throughput increased to 1,865 RPS, representing a 45.7% improvement. Database CPU utilization was reduced from 81% to 58%, while the cache hit rate remained above 93% during peak traffic. In addition, slow SQL statements were reduced by 61% after index optimization and query restructuring. The results demonstrate that combining multi-level caching with efficient database access strategies can significantly improve backend performance and maintain stable service quality under high-concurrency workloads. These findings provide practical guidance for designing scalable and reliable backend systems in modern web applications.