Improving Service Availabiliy with Rule-Based Adaptation

Abstract

Self-adaptive software systems change their deployment and configuration to address changing user behavior and workloads. Such systems follow a MAPEK approach to observe and analyze the system, and subsequently plan and execute changes. They use operations, like the (de-)replication and migration of components, to reconfigure the system. During an adaptation, some services might become unavailable when services are replicated or migrated arbitrarily. This can cause interruptions to cross service transactions and temporary service malfunctions. While some E-commerce platforms consider this acceptable, it is irritating to the consumer. In case of safety critical systems, like medication systems, the system must be serviceable during the transition from the old to the new configuration. In this paper, we present a rule-based approach for adaptation actions. Our approach allows to address adaptation constraints on a abstract level and decouples the constrains from setup scripts often used in container environments. Furthermore, we evaluated the feasibility of our approach and illustrate its ability to adapt a component based web system safely.