近年來隨著物聯網 (IoT) 與人工智慧 (AI) 技術的蓬勃發展，智慧物聯網 (AIoT) 的應用服務也越來越熱門。由於服務對於低網路延遲、高資料隱私等需求的增加，使得邊緣運算相對於雲端運算擁有較為顯著的優勢。 然而，在邊緣運算的環境中，除了運算資源的限制外，運算設備也容易因為能源限制或過熱等環境因素影響而導致運行效率突然降低。而在AIoT的應用服務當中，所執行的運算任務通常需要處理連續的資料流，且任務可能會有著前後執行的關聯，因此當運算設備遇到突發的運算效能下降而使部分任務無法符合服務品質 (QoS) 的限制時，則可能會造成後續任務的延誤或停滯。 為了解決上述問題，本研究將在邊緣智慧物聯網的應用架構下，針對資料流導向任務的工作流程提出自動化的負載平衡解決方案。在任務因運算設備的效能變動而導致過載時，可以自動化的進行偵測與判斷，並將過載的任務轉移至合適的節點，恢復整體工作流程的運行效率。;In recent years, due to the vigorous development of Internet of Things (IoT) and Artificial Intelligence (AI) technologies, the applications of Artificial Intelligence of Things (AIoT) has become more and more popular. Because of the increasing demand for such as the low latency and high data privacy of services, edge computing has more prominent advantages than cloud computing. However, in the edge computing environment, except for the computing resources constraint, the computing efficiency of computing devices will be affected by energy constraints or physical factors such as temperature. Moreover, the computing jobs of AIoT usually deal with continuous dataflows, and maybe context-dependent. As a result, the decrease of computing efficiency may lead to the quality of service (QoS) of jobs unaffordable, moreover, causing the workflow stagnant. To solve the problem, in this paper, we propose the automatic load balancing solution for dataflow-oriented jobs in edge computing. Automatically detect the overloaded jobs and transfer them to appropriate edge servers to resume the calculations and make the workflow back to the normal state.
