Investigating the Impact of Congestion Control Algorithms on Edge-Cloud Continuum

Edge-Cloud Continuum (ECC) is an architecture combining the concepts of Cloud and Edge Computing to improve the performance of distributed services. This combination is necessary to improve Quality-of-Service (QoS) indicators, related to latency, processing time, and energy consumption. Essentially, services are simultaneously hosted by edge and cloud servers, distributing the workload based on user's expectations. Although revolutionary, the ECC architecture is dependent on the interconnection networks, inheriting its research challenges related to the management and shared resources. Even applying optimized provisioning policies (scheduling and allocation), the end-to-end performance of distributed applications remains dependent on Transmission Control Protocol (TCP)'s internal algorithms. We investigate the impact of the congestion control algorithms Cubic, Reno and Bottleneck Bandwidth and Round-trip propagation time (BBR) when used for supporting ECC applications. Our experimental analysis comprises two scenarios: (i) the execution of n-layer application using resources distributed atop edge and cloud providers, and configured with Cubic and Reno algorithms; and (ii) a scenario demonstrating the benefits of using BBR in ECC architectures. In summary, the analyzes demonstrate that the congestion control configuration can improve application performance in ECC.