CARE-Mid: Context-Aware IoT Middleware for Real-Time Resource Management in Smart Environments

Abstract

Smart settings, such intelligent buildings, campuses, and hospitals, rely on a lot of different IoT devices that are always sending and receiving data and control tasks. Most middleware solutions today only focus on either connection or static service orchestration. They don't know much about the surrounding context, future workload, or Quality-of-Service (QoS) limits. This often results in poor task placement, wasteful cloud offloading, and frequent Service Level Agreement (SLA) violations when traffic is high. This article puts forth CARE-Mid, a context-aware IoT middleware designed for real-time resource management across both edge and cloud levels. CARE-Mid combines four important features: (i) getting semantic context and evaluating Quality-of-Context (QoC), (ii) predicting workload and resource demand based on context, (iii) scheduling multiple objectives that optimise latency, energy, and SLA compliance at the same time, and (iv) adaptive priority control based on current utility and recent violations. The framework is set up as a microservice-based platform on an edge-cloud testbed and tested with a genuine smart-building dataset. Experimental results indicate that, in comparison to a standard baseline middleware, CARE-Mid decreases average end-to-end latency by 20–30%, enhances the deadline-hit rate to around 97%, and minimises SLA violations by over 50% during heavy load conditions. A context-aware hybrid placement technique, on the other hand, uses 20–25% less energy and costs 20–25% less to run than cloud-only execution. Ablation experiments show that context modelling, prediction, and adaptive scheduling all play a big role in these improvements.