Abstract:

The increasing demand for highly responsive, scalable, and real-time web applications has exposed fundamental performance limitations in traditional full-stack architectures that rely on synchronous, blocking communication models. Such systems often suffer from elevated latency, limited throughput scalability, inefficient resource utilization, and degraded frontend responsiveness under high concurrency. Reactive programming, characterised by asynchronous execution, non-blocking I/O, and event-driven data streams, has emerged as a promising alternative; however, its performance benefits have largely been studied in isolation at either the frontend or backend layer. Comprehensive empirical evaluations of end-to-end reactive integration across the full stack remain limited.

This study presents a comparative experimental evaluation of a traditional synchronous full-stack application and a fully reactive full-stack application designed with coordinated reactivity between frontend and backend components. Both systems were implemented with identical business logic, data models, and user workflows, differing only in their architectural paradigms. The applications were subjected to controlled workloads encompassing read-heavy, write-heavy, and mixed request patterns under low, medium, and high concurrency levels. Performance was assessed using a multidimensional set of metrics, including average and tail latency (P95, P99), throughput, CPU and memory utilisation, thread consumption, frontend rendering behaviour, UI update latency, and system stability indicators such as error rates and backpressure effectiveness. Experimental results demonstrate that the reactive full-stack architecture significantly outperforms the traditional system across all major performance dimensions. Under high concurrency, the reactive system achieved a 55–65% reduction in average response latency and approximately 70% reduction in tail latency, while sustaining more than 120% higher throughput. Resource utilisation analysis revealed substantially lower CPU and memory consumption and a dramatic reduction in active thread count due to non-blocking execution. Frontend performance measurements showed faster time to first contentful paint, reduced UI update latency, and smoother rendering under real-time data streams. Furthermore, the reactive system maintained low error rates and stable operation under peak load through effective backpressure-aware communication, whereas the traditional system exhibited frequent timeouts and failures. The findings confirm that performance optimisation in modern web applications is fundamentally a cross-layer concern and that isolated adoption of reactive techniques is insufficient to realise their full benefits. By empirically demonstrating the advantages of coordinated reactive frontend–backend integration, this study provides practical architectural insights and design guidelines for building scalable, resilient, and high-performance full-stack applications suitable for real-time and high-concurrency environments.

Keywords:

Full-stack web applications; performance optimization; reactive programming; asynchronous processing; non-blocking I/O; frontend responsiveness; backend scalability; microservices architecture; event-driven systems; real-time data streaming; backpressure management; high-concurrency workloads

Aim/Objectives

Methods