隨著AIoT設備逐漸變多，設備管理與佈署也變多，顯得更不容易管理且複雜，本研究以輕量級微服務架構，以人臉辨識服務作為範例，其識別內容包含性別、年齡、情緒、身分等，應用於AIoT閘道器上並進行拓展佈署與更換模型應用服務，只需要透過一個管理端即可佈署應用與資源分配至所有節點，讓服務可以快速應用及提升服務效率。以往用微服務架構設定過於複雜且架構應用服務資源損耗較高，本研究以輕量級微服務架構作為服務框架，在於管理端記憶體減少約25%，節點端記憶體減少約36%，將記憶體資源可用於更多應用服務。本研究實驗成果可應用於各個IoT、AIoT管理服務及拓展應用服務。;With the gradual increase in the number of AIoT devices, device management and deployment also become more and more difficult to manage and complex, this study uses a lightweight microservice architecture with face recognition service as an example, in which the recognition content includes gender, age, emotion, identity, etc., applied to AIoT gateways and extended deployment and replacement of model application services, only through a single management terminal can deploy applications and resources Allocate to all nodes, so that services can be applied quickly and improve service efficiency. In this study, the lightweight microservice architecture is used as the service framework, which reduces the memory of the management side by 25% and the memory of the node side by 36%, making the memory resources available for more application services. The results of this study can be applied to various IoT and AIoT management services and extended application services.
