Description
The project studies the elasticity of microservices-based applications deployed across the cloud-edge continuum.
Modern applications are often divided into loosely coupled microservices. This architecture supports modularity and more precise resource allocation, but it also makes decision-making more complex: congestion in one microservice may affect other components and degrade the overall application performance.
The main goal is to develop an elasticity controller capable of deciding, in real time, which resources should be assigned to each microservice in order to minimize infrastructure usage while preserving quality-of-service requirements.
The proposal combines Layered Queueing Networks (LQN), useful for representing dependencies and blocking behavior in distributed systems, with Graph Neural Networks (GNN), which are well suited to learning over relational structures and generalizing across applications.