在有著行動邊緣運算的雲端無線接取網路架構下的轉發策略

Abstract

隨著低延遲應用程式的興起，我們需要一個新的行動網路架構來支援此類的應用程式，有著行動邊緣運算的雲端無線接取網路可以提供低延遲的特性，將基頻處理和行動邊緣運算應用程式結合在一起可以讓我們更有效率的使用運算資源，而基頻處理和行動邊緣運算應用程式的結合是以通用處理器為基礎，在此篇論文中，我們提出數學模型來分析通用處理器平台。首先，我們把行動邊緣運算應用程式塑造模型，比較先入先出、處理器共享和擴增型處理器共享的模型。下一步，我們加入基頻處理，並分析基頻處理和行動邊緣運算應用程式的資源分配，除了數學模型和模擬結果外，我們還提出一個轉發策略演算法，依照目前的資料量來動態安排資料路徑，這個演算法也有實驗在實際的測試平台上，驗證結果顯示，我們能依照應用程式的需求去控制應用程式的等待時間。

With the arising of the low-latency applications, we need a new mobile network architecture to support this kind of applications. The cloud radio access network (C-RAN) with mobile edge computing (MEC) architecture can provide the low-latency attribute. The combination of the baseband processing and the MEC application can let us use the computing resource more efficiently. The foundation of combining the baseband processing and the MEC application is the general purpose processor (GPP). In this paper, We propose the mathematical model to analyze the GPP platform. First, we model the MEC stage by comparing the first-in-first-out (FIFO), processor-sharing (PS) and generalized-processor-sharing (GPS). Next, we take the baseband stage into our consideration and analyze the resource allocation between the baseband processing and the MEC application. In addition to the mathematical model and the simulation result, we propose the forwarding policy algorithm to dynamically arrange the data path according to the current traffic. The algorithm is implemented on the real testbed. The demonstration shows that we can control the waiting time corresponding to the application requirement.