Declarative Deployment Planning for Green Pulverised Collective Computational Systems

To promote non-functional goals (e.g., energy efficiency and reactivity) in system implementations, multiple strategies can be adopted, including the partitioning of distributed applications and the smart deployment of the resulting sub-components across the edge-cloud continuum. Within the aggregate computing approach to collective adaptive systems engineering (e.g., IoT ecosystems and robot swarms), the pulverisation model of partitioning and deployment works by splitting the collective computation into device computation rounds and in turn the device computation round in terms of five components: sensing, actuation, behaviour, state, and communication components. Previous research has investigated how different deployments of pulverised systems can provide different trade-offs involving performance and efficiency, with methodologies and simulation tools to carry out the comparison. However, there is still no contribution about the generation or search of effective deployments in the first place. To address this gap, this work introduces Declarative Deployment Planning for Pulverised Systems (DePPS), an approach and toolchain based on simulation and a Prolog-based planner to guide the search of candidate deployments of pulverised systems. The benefits of the approach lie in its declarativity, modularity, scalability, and amenability for continuous reasoning of deployment alternatives. We exercise the approach with synthetic experiments and find out that we can achieve “greener” deployments (i.e., with low energy consumption and carbon footprint) while preserving good latencies compared to uniform peer-to-peer deployments.