Latency-Aware RDMSim: Enabling the Investigation of Latency in Self-Adaptation for the Case of Remote Data Mirroring

Self-adaptive systems are able to adapt themselves according to changing contextual conditions to ensure a set of predefined objectives (e.g., certain non-functional requirements like reliability) is reached. For this, they perform adaptation actions which have an effect on the objectives. But, this effect is not reached immediately. Instead, there is a latency between performing an adaptation action and the effect on the objective. In this paper, we present an exemplar which allows to investigate this latency of adaptation actions for a remote data mirroring system (RDMs) based on a previous exemplar, which wasn't latency-aware. The purpose of an RDM is to replicate a data package among multiple locations (called mirrors) to protect against data loss and unavailability. We present a simulation framework for RDM systems, which offers various adaptation actions: changing the number of mirrors, changing the number of links per mirror and changing the topology of the network. The framework is able to predict the effect on three selected non-functional requirements (cost, performance and reliability) and how long a reconfiguration will take, i.e., the latency. We show how to use this information to construct a simple, exemplary optimiser.