Delivering Resilient Virtualized Services in Smart Grid Environments

Cloud computing is currently a hot topic and the ICT industry is witnessing migration of legacy application to the cloud. Its advantages such as theoretical limitless scalability and high reliability of deployed services looks appealing to every industry that needs an IT infrastructure. The electrical industry is not an exception, monitoring and control systems profoundly rely on a resilient and reliable IT infrastructure. In this industry, latency matters since control systems rely on real time data to work. It is therefore not acceptable to work with data that might be already old when it reaches the control systems and the cloud might not be near enough to where measurements are produced to guarantee an efficient control loop. Edge/Fog computing are paradigms that aim at solving these very issues: bringing critical services as near as possible to the edge of the network. Placing services near to the source of data enables earlier analysis and computation, as well as earlier decisions. In this context technologies like Kubernetes looks very promising but need to be extended or supported by other tools to allow workloads to be run on possibly cheap hardware and making possible to manage thousands of nodes or even clusters that are geographically distributed The goal of this thesis work is to present a possible approach to build a resilient infrastructure that is able to run and orchestrate virtualized services on thousands peripheral sites. Multi-cluster solution have been analyzed as well as more specific edge oriented solutions to meet the special needs of the scenario. Particularly, eventualities such as hardware or software failure and network partitioning have been considered as critical and possible issues that the infrastructure should withstand.