透過您的圖書館登入
IP:18.116.23.178
  • 學位論文

設計基於 eBPF 流量感知 5G 核網自動縮放機制於 Kubernetes 架構

Designing an eBPF Traffic-aware 5G Core Network Autoscaling Mechanism based on Kubernetes Architecture

指導教授 : 陳弘明
共同指導教授 : 盧永豐(Yung-Feng Lu)
若您是本文的作者,可授權文章由華藝線上圖書館中協助推廣。

摘要


摘要 隨著第五代行動網路 (5G) 用戶數量逐漸增長,為了改善第四代行動網路的擴充 性及彈性不足情況,3GPP 於 5G 的規範中提出可以利用網路功能虛擬化 (Network Function Virtualization, NFV) 來構建整個 5G 系統。而隨著網路功能虛擬化 (NFV) 與雲 原生 (Cloud Native) 相關的技術越趨成熟,5G 營運商開始嘗試將 5G 核心網路元件部署 於如 Kubernetes 等容器調度管理平台運行。而 5G 核心網路中,負責處理 UE 流量的 UPF,會使用 GTP-U 通訊協定在 5G 核心網路內傳輸資料,並使用 TCP/IP 通訊協定將 資料傳輸至外部資料網路(Data Network)。然而,Kubernetes 所提供的自動化擴展機制, 無法針對進行通訊協定轉換的 UPF 進行有效的自動化擴展。因此無法根據當前 5G 網路 的負載量調控 UPF 的處理能力,以達到對 5G 核心網路的流量達到動態的負載平衡能力 調整。因此,本研究就 5G 網路、雲原生的 Kubernetes 叢集及 eBPF 相關的技術進行研 究,並設計基於 eBPF 之 5G 流量感知與自動擴展系統架構。將處理封包之 eBPF 附加 於 UPF 之虛擬網路介面卡上,以用於將負責使用者資料傳輸 5G 網路元件 UPF 上進行 流量感知與轉發等功能。以此達到對 5G 核心網路的流量進行動態的負載平衡能力調整與觀測。

關鍵字

5G 核心網路 Kubernetes eBPF

並列摘要


Abstract With the increasing number of 5G mobile users, the 3GPP(3rd Generation Partnership Project) has proposed to utilize NFV(Network Function Virtualization) to construct the entire 5G system in the 5G standard to address the limitations of scalability and flexibility of the 4th Generation mobile network. As the technologies related to NFV and Cloud Native become more mature, 5G operators have begun deploying 5G Network Functions onto container orchestration and management platforms such as Kubernetes. The 5G core network relies on User Plane Functions (UPFs) to handle user traffic. These UPFs employ the GTP-U protocol to transmit data within the 5G core network and the TCP/IP protocol to transfer data to external data networks. However, the automated scaling mechanism provided by Kubernetes is not equipped to effectively manage the scaling of UPFs involved in protocol conversion. Consequently, the 5G core network's ability to dynamically load balance traffic is hindered by the inability to adjust UPF processing capacity based on the current 5G network load. This research delves into the integration of eBPF technology within 5G networks, cloud-native Kubernetes clusters, and UPFs to establish an eBPF-based 5G traffic aware and auto-scaling system architecture. By attaching eBPF packet processing to the UPF's virtual network interface card, the system enables traffic awaring and forwarding capabilities for UPFs, the 5G network function that responsible for user data transmission. This approach empowers dynamic load balancing and traffic observation for the 5G Core Network.

並列關鍵字

5G Core Network Kubernetes eBPF

參考文獻


參考文獻
[1] Ericsson. (2023, November). Ericsson Mobility Report November 2023.
https://www.ericsson.com/4ae12c/assets/local/reports - papers /mobility-
rep or t/ do cu m en ts / 202 3/ er i cs s o n - mob i li t y - r epo rt - n ov em be r - 2 023 .pdf
[2] 3GPP. (2018). Management and orchestration; network policy management for mobile

延伸閱讀