• cloud
• finite state machines
• horizontal scaling
• management protocols
• replicas

Modern applications are composed of multiple heterogeneous components, whose management must be orchestrated by taking into account inter-component dependencies and potential failures. In this thesis we present our extensions to the true concurrent fault-aware management protocols to support horizontal scaling of a multi-component application, i.e., how to properly manage multiple copies of the same component concurrently active at runtime. We also illustrate how such protocols can be composed to automatically derive the replica-aware management behaviour of a multi-component application, and how they enable interesting analyses, like recovering a component that got stuck because of an unexpected fault, or checking whether a management plan is valid or not. We also present a technique to solve the problem of redundancy using replica controllers with health probes.