YUMA – An AI Planning Agent for Composing IT Services from Infrastructure-as-Code Specifications

Abstract

Infrastructure-as-code enables cloud architects to automate IT service delivery by specifying IT services through machine-readable definition files. To allow for a reusability of the infrastructure-as-code specifications, cloud architects specify IT services as compositions of sub-processes. As the AI planning agents for automated IT service composition proposed by prior research fall short in the infrastructure-as-code context, we design a search-based problem-solving agent named YUMA according to a design science research process to fill this research gap. YUMA holds a search tree reflecting the state space and transition model. It includes an algorithm for building the search tree and two algorithms for determining the minimum composition plan. The underlying IT service composition problem is explicated for the infrastructure-as-code context and formulated as a search problem. The results of the demonstration and evaluation show that YUMA fulfills the requirements necessary to solve this problem and digitizes an important task of cloud architects.