考量多服務優先權於軟體定義網路之資源協作最佳化演算法

Abstract

5G軟體定義網路被視為第五代行動通訊網路 (5G) 的關鍵技術，預期可以實現集中式的無線接取網路、網路切片和網路功能虛擬化等技術。並且透過軟體定義網路控制器與新的網路協定，能夠更有效的分配資源，使得未來網路架構更有彈性、效率及可靠性。 然而，當網路切片形成一種服務，目前仍缺乏提供垂直產業或市場對於不同服務提供不同對應的優先級概念。而網路實體層中的分離問題和端到端服務需求的延遲問題仍舊需要被解決。因此在本篇論文中，提出了優先級概念來分配多種服務要求在網路切片中的路徑選擇與優先級設定並計算其延遲，以期望每種服務需求都能獲得延遲要求的滿足。 這個複雜的問題被我們進一步設計為數學模型，目標是最小化所有端到端服務要求的平均延遲。並且利用拉格朗日鬆弛法來解決該模型，並且發展了以拉格朗日鬆弛法為基礎的演算法來求可行解。最後，透過了實驗情境來證明我們所提出演算法的效果，優先級概念的幫助與延遲的降低都證明了其能力，可有效提供系統維運之參考。

The 5G software-defined network is regarded as the key technology of the fifth generation mobile communication network (5G), and it is expected to realize technologies such as centralized wireless access network, network slicing, and network function virtualization. And through software defined network controller and OpenFlow protocol, resources can be allocated more efficiently, making future network architecture more flexible, efficient, and reliable. However, when network slicing become a service, there is still a lack of the concept of providing vertical industries or segments with different corresponding priorities for different services. The isolation problem in the physical layer of the network and the delay of end-to-end service requirements still need to be solved. Therefore, in this thesis, the concept of priority is proposed to allocate multiple service requirements, and the delay is calculated to expect that the delay tolerance of each service requirement can be met. This complex problem is designed as a mathematical model with the goal of minimizing the average latency of all end-to-end service requirements. The Lagrangian relaxation method was used to solve the model, and an algorithm based on the Lagrangian relaxation was developed to find a feasible solution. Finally, through experimental scenarios to prove the effectiveness of our proposed algorithm.