Fault Tolerant Robust Adaptive Workload Orchestration in Pure Edge Computing

Pure Edge Computing (PEC) emerges as a solution to meet the increasing demand for time-sensitive and data-driven applications by bringing cloud applications and services closer to end users. Nevertheless, the mobile nature of edge devices and their inherent computing limitations pose several challenges in supporting urgent and computationally intensive tasks. Tasks become worthless and can lead to serious safety issues if they fail to meet the deadline. Furthermore, device crashes can trigger significant system failures. Hence, it is crucial for such systems to handle failures and be fault tolerant to maximise the completion of latency-sensitive tasks in accordance with available resources. This paper proposes a Fault-Tolerant Robust Adaptive Workload Orchestration (FTR-AdWOrch) that handles uncertainties caused by resource failures in dynamic environments. Resource prioritization and a reallocation mechanisms are designed to minimize the number of missed deadlines for task completion and data loss. We propose a fault-tolerant orchestrator that utilises a backup node to prevent any potential orchestrator node failure, which plays a vital role in workload management and the system performance. The simulation results show that FTR-AdWOrch can minimize missed deadlines of urgent tasks while minimizing the data loss of lower priority tasks regardless of dynamic conditions.