Scalable Multi-Algorithm Distributed Rate Limiting: A Study on Redis-Lua Atomicity, Resilience4j Fallbacks, and High-Fidelity Observability

As modern software architectures shift toward highly decentralized microservice ecosystems, the challenge of maintaining service stability under unpredictable traffic patterns has become a central concern. This research presents a production-grade Distributed Rate Limiting framework implemented using Spring Boot and Java. The primary technical contribution is the development of a dual-layered enforcement strategy that leverages server-side Lua scripts for atomic state transitions within a Redis cluster [1]. We evaluate the performance of four core algorithms: Token Bucket [2], Leaky Bucket [3], Fixed Window, and a memory-resident Sliding Window. The system features a first-class observability stack involving Micrometer and Prometheus [5], [6]. Our analysis demonstrates high availability through Resilience4j-based local in-memory fallbacks [4].