Introducing Automatic Discovery Mechanisms for the Computing Continuum

The optimization of computing resources at the edge, combined with cloud-based computational power, has gained significant attention in recent years under the concept of the computing continuum. Within this context, the Flexible, scaLable, secUre, and decentralIseD Operating System (FLUIDOS) project has emerged as an implementation of this paradigm, enabling seamless resource management across edge and cloud environments. Among its potential applications are optimizing resource utilization for a fleet of robots, reducing power consumption while enabling intensive computations, and facilitating interaction between connected vehicles and infrastructure. A challenge in this landscape is the discovery of available computational resources. The scope of this thesis is the design and implementation of an evaluation framework to benchmark, analyze and compare different discovery protocols within the FLUIDOS ecosystem. The framework provides a structured approach by defining network condition requirements, setting up a testbed with defined and repeatable conditions, and implementing a benchmarking tool compatible with the FLUIDOS architecture, capable of automating tests to ensure statistical significance. The goal is to compare various protocol alternatives based on meaningful performance metrics such as discovery time across varying network conditions and protocol behavior at scale. Experimental results highlight the distinct behaviors of discovery protocols based on multicast and distributed hash table (DHT) overlay networks. The analysis reveals significant differences in discovery time, scalability, and inherent networking limitations, providing insights into the trade-offs between these approaches. These findings contribute to a deeper understanding of resource discovery mechanisms and offer a foundation for further development of discovery protocols in FLUIDOS, giving users the ability to choose the most suitable protocol for their specific use case.