Artificial Intelligence Workload Allocation Method for Edge-Cloud Computing Environment

Edge-cloud computing paradigms increase the quality of experience (QoE) for real-time artificial intelligence (AI) applications by offering many benefits, such as reducing the response time. Efficient workload allocation methods between edge and cloud are essential to meet the AI applications demand, hence increasing the QoE. Workload allocation is known as a nondeterministic polynomial (NP) time problem (NP-hard problem), which is difficult to resolve using conventional techniques like heuristic algorithms. The difficulties derive from the fact that several factors influence the allocation process, such as application characteristics, network conditions, and resource-related issues. In this work, an AI system is proposed based on deploying different AI applications on edge and cloud servers, and then managing the workload based on AI techniques. The presented system considers the AI application that has slow or static movement in the generated task level (end device), such as smart devices in the Intensive Care Unit (ICU) of the hospital or monitoring cameras. The proposed system is called Workload Allocation for Fixed Application (WAFA). The proposed WAFA system consists of three AI applications deployed on edge servers and AWS cloud servers. The allocation decision model was trained using a pre-generated dataset based on several features. The proposed WAFA method outperforms the random, edge, and cloud methods by 41.47%, 77.96%, and 38.10%, respectively.