Congestion Control with Service Mesh for Edge Cloud Traffic

Abstract: The Mobile Edge Cloud aims to expand the cloud network by making it distributed and closer to the end user. However, traffic management becomes a complex and unresolved problem, because the edge cloud needs to consider multiple locations with different computational resources. This thesis contributes to the solution of this problem by attempting to align cloud and mobile network technology domains, by using the mechanism of Quality of Service assurance from the mobile network to improve the cloud performance, as well as tools like the Service Mesh, which comes from the central cloud. In this thesis, we aim to extend the implementation of the Service Mesh, using the concept of Quality of Service, and provide optimized congestion control for applications hosted in a mobile edge cloud. For this, we design, implement, and evaluate two heuristic-based controllers that act upon the Service Mesh, using resource and traffic metrics to limit the rate of requests based on their priority. The first controller is called Continuous Rate Limiting, where the rate limit configuration remains the same if there is no more congestion, and the second one is called Conditional Rate Limiting, where the controller resets the rate limit configuration if there is no congestion after a certain amount of time. Our research concludes that both controllers reduce resource consumption and response time under a congestion scenario. For instance, the resource consumption decreases down to 18.25% when the Continuous Rate Limiting controller is implemented. The average response time also is reduced down to 44.96% with the same controller. All these results are obtained using the same traffic traces, hence future research is needed for further validation of this thesis. Future research can also use this thesis as a starting point for more advanced traffic management strategies for mobile edge cloud.