A middleware approach to reducing communication latency between distributed microservices, through topology-aware, node-affinity routing.

The master’s dissertation "A Middleware Approach to Microservices Latency Reduction" addresses one of the recurring bottlenecks of cloud-native architectures: as applications are decomposed into distributed microservices spread across regions, providers, and on-premises infrastructure, service-to-service communication becomes a critical source of latency, inter-regional traffic, and routing complexity. Building on a literature review of cloud and distributed computing, microservices architectures, and performance evaluation, the work proposes a lightweight, portable, and vendor-independent routing middleware designed to sit between clients and internal microservices.

The proposed middleware performs topology-aware routing: it resolves the hostname of the node handling a request and forwards the entire microservice pipeline to that same node, preserving processing affinity and avoiding unnecessary inter-node and inter-regional communication. The solution was implemented and evaluated in a Docker Swarm environment deployed across Amazon Web Services, Microsoft Azure, and an on-premises cluster, comparing scenarios with and without the middleware under different load profiles. The results show consistent latency reductions across all three environments, positioning the middleware as a lightweight alternative to traditional API Gateways and Service Meshes.