⾏動邊際運算：多營運商無線電接入網路分享、多租戶切片以及服務鏈路由

Mobile Edge Computing: Multi-operator RAN Sharing, Multi-tenant Slicing, and Service Function Chain Routing

簡旭彤(Hsu-Tung Chien)

指導教授 ： 林盈達

國立交通大學/資訊學院/資訊科學與工程研究所/博士(2019年)

摘要

為了滿足5G網絡中三種服務類型的需求，移動邊緣運算 (MEC) 解決辦法被提出。此外，為了實現靈活且環境獨立的各種服務部署，將基於虛擬化技術的網絡切片引入5G系統。多租戶端到端切片包括Radio Access Network (RAN) 和Core Network的2層結構，在此架構中有computing和communication兩種資源。每個切片都可以視為一個完整的MEC系統，此系統支援Service Function Chain (SFC)的部署。為了完成此複雜系統的開發，我們需要幾個關鍵模組，包括RAN共享，切片管理結合請求轉換和容量分配，服務鏈路由和MEC管理器。在論文中，我們提出了一種透明的RAN共享方法，即RAN代理，具有公平協議，是一個利用Soft partition 與 blocking和Dropping (SBD)概念的方法。 SBD實現了0.997的運營商間公平性。此外，當共享BS未充分利用時，SBD將blocking rate從硬分區方案下的35％降低到幾乎0％，而控制dropping rate在5％以下。但在啟動scale down 功能後，可以將dropping rate降低到幾乎0％。關於切片管理，提出了聯合邊緣和中央資源切片器 (JECRS)並且使用開源工具將其實現。實驗結果表明，該框架成功地隔離了切片之間的5G資源。此外，基於此框架，提出了一種2層資源分配算法 ，稱之在延遲限制下上層優先過度供應預防演算法（UFLOP），此演算法以最小化“過度供應”比率的方式調整運算資源和網路資元分配在滿足租戶的延遲限制下。 實驗數據說明了UFLOP成功且正確的分配上層和邊緣之間的資源比率，實現了接近零的過度配置比率，同時仍滿足延遲要求。結果表明，增強型移動寬帶（eMBB），超可靠低延遲（URLLC）和大規模機器類型連接（mMTC）應用的最佳資源分配比分別為10：0, 1.5：8.5和7.8：2.2。至於路由，多站點最短端到端延遲第一採用參考鏈路和節點狀態路由（MM-LNSR）是一種用於有線/無線網絡中SFC的路由計算算法。透過修改的Dijkstra和旅行商問題算法，讓這樣種演算法能不只考慮鏈路的情況同時也考慮運算節點狀態並試圖找出滿足服務的端到端延遲約束的路由。實驗表明，在2-stop情況下，通過另外考慮用於計算VNF的延遲的節點狀態的路由度量，未滿足路徑的百分比顯著改善，減少75％（從80％減少到5％）。並且可用路徑的錯誤率降低了94％（94％至0％），96％（99％至3％）和93％（100％至7％）分別是在一站式，兩站式和三站式的情況下考慮鏈路可靠性和VNF的路由度量時。最後， 提出的MEC平台部署解決方案於4G LTE網絡中採用了中間盒方法。它符合標準且對現有移動網絡組件可以兼容，並通過託管應用服務器而為移動用戶提供MEC服務。我們通過基於開源OpenAirInterface移動網路平台的實作證實了它的可行性。綜上所述，本文提出了許多滿足5G網絡服務類型需求的解決方案。

關鍵字

LTE ； 5G ； 網絡切片 ； RAN共享 ； 移動邊緣運算 ； 服務功能鏈 ； 路由 ； 容量分配 ； 計算資源 ； 通信資源

並列摘要

To meet the needs of three service types in 5G networks, mobile edge computing (MEC) has emerged. Furthermore, to achieve flexible and isolated deployment of diverse services, network slicing into virtualized platforms has been brought into the 5G systems. An end-to-end slice consists of both computing and communication resources deployed across the 2-tier structure of access and core networks. Each slice can be viewed as a complete 5G MEC system supporting service orchestration and chaining. To complete the development of the platform, we need several advanced components, including RAN sharing, slicing management with request translation and capacity allocation, service chain routing, and MEC manager. In this dissertation, we proposed a transparent RAN sharing method which is RAN Proxy with a fairness protocol a Soft-partition with Blocking and Dropping (SBD). SBD achieves an inter-operator fairness of 0.997. Furthermore, SBD reduces the blocking rate from 35% under a hard partition scheme to almost 0% when the shared base station is under-utilized, whereas controlling the dropping rate at 5%. Notably, the dropping rate can be reduced to almost 0% using a newly proposed bandwidth scale down procedure. About the slicing management, the proposed Joint Edge and Central Resource Slicer (JECRS) framework is implemented using open source tools. The experimental results show that the framework successfully isolates the 5G resources between slices. Moreover, relied on the framework, Upper-tier First with Latency-bounded Over-provisioning Prevention (UFLOP), a 2-tier resource allocation algorithm, is proposed to adjust the capacity and traffic allocation in such a way to minimize "over-provisioning ratio" while still satisfying the latency constraints of the tenants. UFLOP successfully determines the traffic allocation ratio between the central office and the edge, which achieves an over-provisioning ratio close to zero while still meeting the latency requirements. The results suggest optimal resource allocation ratios of 10:0, 1.5:8.5 and 7.8:2.2 for the Enhanced Mobile Broadband (eMBB), Ultra-Reliable Low Latency (URLLC), and massive Machine Type Connection (mMTC) applications, respectively. For routing, we proposed a Multi-stop Multi-path Link and Node State Routing (MM-LNSR), a route calculation algorithm for the SFC in wired/wireless networks with modified Dijkstra's and Traveling Salesman Problem algorithms, considering the proposed link in addition node states and trying satisfy the end-to-end latency constraints of services. The experiments show a significant improvement of the percentage of unsatisfied paths, reduced 75% (from 80% to 5 %) by additionally considering the proposed routing metric of the node state for computing delay of VNFs in 2-stop case, and the failure rates of available paths are reduced 94% (94% to 0%), 96% (99% to 3%), and 93% (100% to 7%) of 1-stop, 2-stop, and 3-stop cases, respectively, by considering both proposed routing metrics for reliability of link and VNF. Finally, the proposed MEC platform deployment solution in 4G LTE networks is using a middlebox approach. It is standard-compliant and transparent to existing cellular network components, and enables the MEC service for mobile users by hosting application servers. We have confirmed its viability through a prototype based on the open-source OpenAirInterface cellular platform. To sum up, the dissertation proposes solutions to meet the needs of service types in 5G networks.

並列關鍵字

LTE ； 5G ； network slicing ； RAN sharing ； mobile edge computing ； service function chain ； routing ； capacity allocation ； computing resource ； communication resource

參考文獻

[1] ETSI, “ABOUT ETSI”, [Online]. Available: https://www.etsi.org/about. [Accessed: Aug. 2019].

[2] 3GPP, “Study on enhancement of Ultra-Reliable Low-Latency Communication (URLLC) support in the 5G Core network (5GC),” 3GPP specification, 3GPP TR 23.725 V0.3.0, Jul. 2018.

[3] 3GPP, “The path to 5G: as much evolution as revolution”, [Online]. Available: https://www.3gpp.org/news-events/partners-news/1969-mec. [Accessed: Aug. 2019].

[4] 3GPP, “5G-NR workplan for eMBB”, [Online]. Available: http://www.3gpp.org/news-events/3gpp-news/1836-5g_nr_workplan. [Accessed: Aug. 2019].

[5] ETSI, “Mobile Edge Computing (MEC) Terminology,” ETSI GS MEC 001 V1.1.1, March 2016.