Bringing Aggregate Programming Towards the Cloud

Aggregate Programming (AP) is a paradigm for developing applications that execute on a fully distributed network of communicating, resource-constrained, spatially-situated nodes (e.g., drones, wireless sensors, etc.). In this paper, we address running an AP application on a high-performance, centralized computer such as the ones available in a cloud environment. As a proof of concept, we present preliminary results on the computation of graph statistics for centralised data sets, by extending FCPP, a C++ library implementing AP. This: (i) opens the way to the application of the AP paradigm to problems on large centralised graph-based data structures, enabling massive parallelisation across multiple machines, dynamically joining and leaving the computation; and (ii) represents a first step towards developing collective adaptive systems where computations dynamically move across the IoT/edge/fog/cloud continuum, based on mutable conditions such as availability of resources and network infrastructures.