PISCIS: A Constraint-Based Planner for Self-adaptive Systems

The dynamic nature of current software execution environments demands a reliable and prompt response to guarantee quality at runtime. Self-adaptive software systems contain facilities that allow for an autonomous and continuous adaptation towards such end. The facilities are, overall, structured in accordance with IBM's MAPE-K reference model. IBM's MAPE-K reference model comprises a Monitor and an Analyzer to sense and interpret context data, a Planner and an Executor to create and apply structural adaptation plans, and a Knowledge Manager to manage and share relevant information. The aim of this paper is to present Piscis, our Planner element based on the principles of a constraint satisfaction problem (CSP) and the MAPE-K reference model. We tested Piscis, within a controlled environment made up of a set of Eclipse plug-ins, and checked how it dynamically creates structural adaptation plans following the MAPE-K reference model. Piscis is a constraint-based planner within the Shift framework, based on the MAPE-K reference model. We have verified its applicability using a running example set within an adaptation scenario.