Network connectivity and observability in multicluster environments

Nowadays cloud computing is becoming more important and the current trend is to engineer the new web applications to be cloud-native, to split up the application into micro-services, each one containerized and deployed as an independent part of a bigger application. A technology that broke through the cloud market is Kubernetes, a project that allows orchestrating containers in a cloud environment, creating an abstraction layer that hides the physical machines, allowing to manage them as a single entity. However, a new trend in the last years is to extend this abstraction to a group of clusters to create another abstraction layer to allow managing them like a single entity. This approach is called multi-clusters, a scenario where multiple independent clusters communicate and create a federation allowing to share resources between them and to exchange workloads. The advantage of a multi-clusters approach is the possibility to share the cluster's resources with others to optimize the usage of the latter. It also allows companies no longer depend on a specific cluster or cloud provider. The goal of this thesis is to study how to interconnect more Kubernetes clusters and allows their microservices to communicate transparently. For this purpose 2 designs have been produced. These designs present a network model to enable communication with minimal dependencies and without the need for specific configurations on clusters. To test the new design in a real environment, the thesis design focuses on Liqo, an open-source project started at Politecnico di Torino. Each design has been implemented and validated. Finally, the 2 designs have been compared to evaluate which is the best. Another goal of the thesis is to provide the user with a simple way to monitor the status of the network and to keep track of the performance of the connectivity towards federated clusters. These observability features have been designed to expose information to other microservices.