Lightweight cloud service composition and orchestration for edge intelligence

The rapid growth of IoT and edge computing has led to a surging demand for intelligent services at the network edge. Deploying and orchestrating such services in edge-cloud environments face challenges due to edge devices’ heterogeneity, dynamicity, and resource constraints. We propose a lightweight cloud service composition and orchestration framework for edge intelligence applications, leveraging container technologies for efficient deployment of data-intensive services across heterogeneous edge nodes and cloud servers. A data-driven container scheduling algorithm, DICS-OPT, is introduced, utilizing multi-criteria decision-making and weight factor optimization to optimize task completion performance, resource utilization, and cloud cost savings. Extensive experiments demonstrate the effectiveness of the proposed framework in handling diverse edge intelligence workloads. DICS-OPT achieves significant improvements, with up to 80% reduction in task completion time, 90% increase in resource utilization, and 70% savings in cloud costs. The framework's scalability limitations are discussed and future research directions are proposed. This work enables developing advanced edge-cloud systems for seamless integration of edge and cloud resources in edge intelligence applications.