Zero-Defect Request Coalescing: A Cooperative-Preemptive Userspace Gateway for Cache Stampede Mitigation across Distributed Cloud, Clinical, and Biosystem Infrastructures

Abstract

Cloud microservices often suffer from database connection-pool exhaustion during sudden traffic spikes, a failure state known as a cache stampede. Traditional mitigation relies heavily on load-shedding, which preserves backend stability by indiscriminately dropping client requests. This paper introduces an application-layer request coalescing gateway that intercepts identical concurrent queries and collapses them into a single-flight database operation. The architecture operates on a single-threaded, cooperative-preemptive event loop and integrates C-based state-machine parsing to completely bypass multi-threaded lock contention. Empirical evaluation across 100,000 synthetic network requests demonstrated a 98.74% mean reduction in backend database queries. Furthermore, the architecture effectively decoupled application throughput from physical socket constraints. Even when host operating system transmission control protocol (TCP) handshake queues degraded to 2,100 milliseconds, the proxy sustained an isolated execution latency between 188 and 599 milliseconds. These results establish userspace coalescing as a highly efficient, non-destructive alternative to traditional edge load-shedding. Crucially, this architecture provides a foundational resilience model for cross-domain infrastructures, securing real-time data access across clinical networks and ensuring uninterrupted telemetry in high-density biosystem environments.