A Lightweight Autonomous and Adaptive Scheduler for Next-Generation Fog Computing Systems

Advanced healthcare apps create latency-sensitive and privacy-sensitive IoMT workload which cannot be effectively managed by faraway Cloud computing alone. LAAS is a lightweight contextual bandit-based scheduler proposed for self-managed adaptive Fog-Cloud computing. LAAS works at the healthcare gateway and hospital Fog level where task criticality classification, workload and unreliability prediction, action filtering based on hard QoS, privacy, energy, and queue constraints, and the execution resource selection based on LinUCB-based contextual bandit policies are executed. Critical tasks are guaranteed to be handled through a fallback mechanism that gives priority to local gateway, hospital Fog, nearby Fog, Fog cluster, and finally to Cloud when needed. Experimental results conducted in an IoMT gateway to Fog computing hardware setup indicate that LAAS minimizes emergency delay time, SLA violations, Cloud fallback, and energy consumption as well as enhancing critical task reliability, privacy-friendly local computation and balanced Fog-node utilization compared to recent IoMT offloading and Fog scheduling methods.