基於單鍊路故障的可生存虛擬網路嵌入

Abstract

隨著物聯網(Internet of Things)設備和需求的快速增加，傳統雲端網路已無法負荷這麼多的應用程式，因此邊緣運算被提出來解決這個問題。而借助網路虛擬化(Network Virtualization)技術，邊緣運算也將變得更加強大。在網路虛擬化中，虛擬網路嵌入(Virtual Network Embedding)是一個重要的挑戰，所謂虛擬網路嵌入是指將虛擬網路請求(Virtual Network Request)嵌入實體網路(Substrate Network)的過程，以滿足虛擬網路的需求和限制。 在虛擬網路嵌入研究中大多假設基礎設施供應商(Infrastructure Providers)提供的資源是恆定不變的，並在此前提下最大化資源利用率並滿足虛擬網路的需求。然而，在實際情況下會因為軟體或硬體故障、惡意攻擊等原因，實體網路中的運算節點或鏈路會出現故障，這將有可能導致虛擬網路服務中斷。為了使虛擬網路服務不中斷，可生存性虛擬網路嵌入(Survivability Virtual Network Embedding)被提出。 可生存性虛擬網路嵌入討論的是確保虛擬網路服務在實體網路資源發生故障時也能持續運行。其問題可以根據實體網路中故障類型不同被分為處理鏈路故障和處理節點故障兩類型。由於鏈路故障發生機率比節點故障高，因此現有的研究主要集中在處理鏈路故障問題而非節點故障問題。在處理鏈路故障問題中，最基本的問題是發生單鏈路故障時該如何使服務不中斷，所以在本論文中，我們提出了一種結合分割和共享路徑的方法，來解決這個問題，以確保一條鏈路發生故障時服務不會中斷。與現有的可生存性虛擬網路嵌入解決方式相比，我們的方法不需保留大量冗餘頻寬，就可以確保虛擬網路的生存性。 我們模擬了小型實體網路及大型實體網路兩種情境，在小型虛擬網路請求嵌入到小型實體網路中，以平均資源預留率而言，我們的方法相比基準線減少0.97。接著我們比較了小型和大型虛擬網路請求分別嵌入到大型實體網路中的結果。在小型虛擬網路請求嵌入到大型實體網路中，以平均資源預留率相比減少1.19。在大型虛擬網路請求嵌入到大型實體網路中，以平均資源預留率相比減少1.24。

With the rapid increase in Internet of Things (IoT) devices and demands, the traditional cloud network is no longer capable of handling such a multitude of applications. Therefore, the edge computing has been introduced to address this issue. Through the utilization of the network virtualization (NV) technology, the edge computing will also become more powerful. In the NV, the virtual network embedding (VNE) stands as a significant challenge. The VNE refers to the process of embedding virtual network requests (VNRs) into the substrate network (SN), aiming to fulfill the demands and constraints of the virtual network (VN). In the research on the VNE, it is often assumed that the resources provided by Infrastructure Providers (InPs) remain constant, and under this premise, resource utilization is maximized to meet the demands of the VNs. However, in real-world, due to factors such as software or hardware failures, malicious attacks, etc., substrate nodes or links within the SN may fail, potentially causing disruptions in the VN services. In order to ensure uninterrupted the VN services, the concept of survivability virtual network embedding (SVNE) has been introduced. The SVNE discusses the assurance of continuous operation of the VN services even when the SN resources experience failures. This issue can be categorized into two types based on the different types of failures within the SN: handling link failures and handling node failures. Due to the higher probability of link failures compared to node failures, existing research mainly focuses on addressing link failure issues rather than node failure issues. In addressing link failure problems, the most fundamental challenge is to prevent service interruption when a single link failure occurs. In this paper, we propose an approach that combines path splitting and sharing to solve this issue, ensuring that service remains uninterrupted in the event of a link failure. Compared to existing methods for the SVNE, our approach doesn't demand the reservation of a significant amount of redundant bandwidth to ensure the survivability of the VNs. We conducted simulations in two scenarios: the small SN and the large SN. In the case of embedding the small VNRs into the small SN, our method reduces the average resource reservation rate by 0.97 compared to the baseline. Next, we compared the results of embedding small and large VNRs into the large SN. For the small VNRs embedded into the large SN, our method reduces the average resource reservation rate by 1.19 compared to the baseline. For the large VNRs embedded into the large SN, our method reduced the average resource reservation rate by 1.24, as compared to the baseline.