現今雲端運算越來越成熟，但隨著5G的到來帶動了IOT的網路通訊設備的增加還有智慧型手機的普及，這些訊務量將會湧入核心網路，造成核心網路的壅塞，所以網路服務商會尋求邊緣網路將計算資源與服務功能下放到邊緣網路，並將低延遲的封包送到邊緣網路來卸載核心網路的負載量。但邊緣網路的資源相較雲端來說有限，所以在部署服務時需要考量資源控管的問題，這裡利用了微服務自主化且低耦合的特性，有了自主化的特性且透過完整成熟的RestAPI就可以直接與其他微服務溝通就不需要花時間等待向集中管理controller的回應，再加上其低耦合的特性，當某個服務負載量大時可以只擴展該服務不需要將整個程式擴展減少資源的占用，故因為這兩個特性可以解決時間敏感且資源有限的低延遲邊緣網路部署問題。基於微服務的特性我們會建立一個微服務部署平台，提供服務功能的部署。本論文也提出基於模糊理論的微服務與Hypervisor的分流偵測機制(Fuzzy-based Service Offloading Detection Mechanism, FSODM)，此機制可以提供微服務還有Hypervisor一個快速且資源使用效率高的自動擴展的資源管理解決方案。;Nowadays, the technology of cloud computing is getting mature. However, with the 5G standard coming, it make the increase of IoT devices and mobile devices more widespread. These traffic will flood into core network, which makes the congestion of core network. Therefore, in order to offload the traffic from core network to edge network, the service providers will deploy their computing resources and services to the edge network, which makes packet low-latency transmitted to the edge network. The resource of edge network is fewer than the resource of cloud computing, it need to consider the problem of use the characteristics of low-couple and autonomy in microservice. With mature RestAPI, we can direct to communicate with other microservice without spending time to wait the response of controller. With the characteristics of low-couple, we can just auto-scale the service to reduce the consumption of resource when the loading of some service get instance. Therefore, with these two characteristics, we can solve the problem of deployment for time-sensitive and constraint of resource. Based on the characteristics of microservice, we proposed a microservice platform and provide the deployment of service. We also proposed a FSODM algorithm which provides a rapid and high resource efficiency auto-scaling resource management solution.
