Design and development of a computing-aware network architecture to support dynamic load balancing across the cloud-edge continuum

Edge Computing is a model that shifts computing infrastructure from centralized data centers to locations closer to end users, enabling low-latency communication, which is required for constrained applications such as those for mobile and IoT devices. However, this shift introduces challenges in traffic routing and resource management, as workloads must be dynamically distributed across multiple edge sites with varying capacities. This thesis proposes a scalable and flexible computing-aware network architecture designed to support dynamic load balancing in edge computing environments. By integrating Segment Routing over IPv6 (SRv6) and Border Gateway Protocol (BGP), the proposed architecture enables dynamic traffic steering based on both computing resources and proximity of data centers, ensuring that service requests are directed not just to the nearest edge site but to the most capable service instance. This is particularly crucial in scenarios where the closest site may lack the necessary computing resources. The effectiveness of this approach is validated on a testbed comprised of GNS3 network with Unix-based routers running FRRouting. The validation results showed that the network architecture successfully managed workloads across multiple service instances, balancing both proximity of the data centers and computing metrics to enhance service delivery and meet the demands of dynamic, high-performance applications. This work contributes to the design of a scalable and flexible network architecture that supports the dynamic demands of edge computing, providing a robust solution for workload distribution across multiple service instances.