Performance-Driven Evaluation of Monolithic and Microservice Architectures for Enterprise Cloud Applications: A Scalability, Latency, and Cost Perspective

As enterprise applications increasingly transition to cloud-based deployments, architectural choices play a decisive role in determining system performance, scalability, and operational cost. This paper presents a performance-engineering-driven comparison of monolithic and microservice architectures for large-scale enterprise cloud applications, with explicit consideration of workload characteristics, specifically database-intensive versus application-intensive processing models. Monolithic systems benefit from in-process communication, predictable latency, and efficient JVM resource utilization, making them suitable for tightly coupled, database-heavy transactional workloads. However, they often encounter scalability and maintainability limitations as functional complexity and workload diversity grow. Microservice architectures, by contrast, enable independent scaling, fault isolation, and deployment flexibility, which are advantageous for application-intensive workloads involving complex business rules, processing flows, and uneven load distribution. These benefits come at the cost of increased latency, JVM overhead, and cloud operational expenses due to distributed execution. This study evaluates both architectures using an enterprise-grade cloud application under varying load conditions, analyzing throughput, tail latency (p90/p99), JVM garbage collection behavior, CPU and memory utilization, database contention, and infrastructure cost. Results indicate that microservices are effective when application-layer processing is the primary bottleneck, while monolithic deployments remain more efficient when the database layer dominates system constraints. The findings emphasize that architectural decisions must be driven by identifying the weakest performance bottleneck and selecting solutions that optimize cost, efficiency, and latency rather than adopting microservices indiscriminately.