Analysis of the Resilience of Monitoring Services in Smart Grid

The benefits of cloud computing, such as scalability and reliability, become established in the electrical industry or also in any company that requires an IT infrastructure. In recent times, more and more industries require monitoring and control system based on a reliable IT infrastructure. Since control systems depend on real-time data to function, it is unacceptable to work with data that may be obsolete before it reaches the control systems. Recently, the traditional electrical grid is facing similar requirements, as the introduction of renewable sources (e.g., solar panels, wind farms, and more) drives the need for enhanced monitoring. These issues are the focus of the edge/fog computing paradigms, which place services close to the new data sources to enable analysis and computation as soon as the data is produced. A scenario where centralised computations shift to the edge is relevant for analyzing the challenges of a geographically distributed infrastructure, such as hardware from various manufacturers as well as physically insecure locations. We also must consider that resources at the edge are less abundant than in a cloud environment and that network partitioning events that isolate one or more sites are a possibility. Hence, each individual site should therefore be resilient to internal failure and able to resist node, control plane, and storage failures. So another crucial factor is the creation of a fault-tolerant system, which is realized using redundant components. In addiction, there was the need to increase the grid observability both to enable a series of mission-critical applications such as congestion control and fault detection and both to balance the power supply and the demand. As a result the need for smart power grids has become more and more prominent because they enable to: use tools to monitor the distributed energy generation; share data from sensors and meters; collect and process data to control the electrical power system. Delivering resilient monitoring and computing services that use real-time data requires strong data resilience. For performing data analysis for statistical meaning or post-incident analysis, historical data durability is definitely essential too. In order to withstand hardware and network failure, mechanisms for performing consistent disk or volume backups and data replication are required. This thesis shows a possible approach to build a resilient database cluster and, consequently, an analysis on resilience in data persistence, on the scalability aspects of the entire infrastructure and on the orchestrator timing involved in a solution that guarantees self-healing, in the face of the issues mentioned above.