Coordinator: Hélène Coulon, IMT Atlantique
Partners: IMT Atlantique (Stack Inria Team), Université d’Orléans (LIFO, LMV Team)
Large distributed software systems (applications or infrastructures) are now ubiquitous, with component-based systems (e.g., service-oriented architectures or microservices) offering a convenient way to structure large systems, in particular distributed systems deployed in the Cloud, in the core, or at the edge of the network. DevOps operations, that include system configurations and reconfigurations, are required to handle various kinds of scenarios such as fault tolerance, scalability, software updates, or various optimizations, etc. Such changes may lead to faults. A study of 597 unplanned outages that affected popular cloud services between 2009 and 2015 found that 16% of them were caused by a system upgrade.
On the one hand, many configuration tools and languages exist in the DevOps community, some of them being specific to the provisioning of resources in Cloud providers, packaging problems, containerized deployments, configuration of applications or infrastructures, etc. The main advantage of these tools is their full integration and adoption in the DevOps community. Their disadvantage is they lack both formal and textual specifications. Moreover, their contours are blurred. On the other hand, many initiatives have been studied in academia to contribute to the deployment, configuration, and reconfiguration of distributed software, bringing improvements such as expressivity, speed, safety, etc. Many come with precise and sometimes formal definitions. However, they lack the breadth of the mainstream DevOps tools.
The goal of For-CoaLa is twofold: (1) understand and bridge the gap between a popular tool from the DevOps community (Ansible) and a tool from academia (Concerto); (2) improve the understanding of these languages based on mechanized formal semantics and develop verified semantic-preserving cross-language transformations.