Skip to Main Content
The configuration and runtime management of distributed systems is often complex due to the presence of a large number of configuration options and dependencies between interacting sub-systems. Inexperienced users usually choose default configurations because they are not aware of the possible configurations and/or their effect on the systems' operation. In doing so, they are unable to take advantage of the potentially wide range of system capabilities. Furthermore, managing inter-dependent sub-systems frequently involves performing a set of actions to get the overall system to the desired final state. In this paper, we propose a new approach for configuring and managing distributed systems based on AI planning. We use a goal-driven, tag-based user interaction paradigm to shield users from the complexities of configuring and managing systems. The key idea behind our approach is to package different configuration options and system management actions into reusable modules that can be automatically composed into workflows based on the user's goals. It also allows capturing the inter-dependencies between different configuration options, management actions and system states. We evaluate our approach in a case study involving three inter-dependent sub-systems. Our initial experiences indicate that this planning-based approach holds great promise in simplifying configuration and management tasks.