基於中轉自治系統之中繼路徑選擇方法

Abstract

Internet 中的各種端點對端點通訊(End-to-end communication)應用正快速的發展中，這類應用的效能跟其下層的Internet架構息息相關，但目前Internet的架構已經使用已久，當初為了單純資料傳遞的架構在現今則大大限制了端點對端點通訊的效能。 以往的研究證明若指定叫原本的端點對端點路徑經過兩端以外的第三點也就是中繼點(Relay Node)，可以大幅改善通訊品質，但目前的方法在尋找優良的中繼點時皆需耗費大量的測量成本(High Probe Overhead)，或是在網路規模增大時無法處理。於是本論文提出一套基於Transit AS概念的中繼路徑選擇機制以尋找較佳的中繼路徑，這個機制的基本想法是利用Transit AS在AS Topology中佔大多數的特性來削減在Relay Path Selection過程中的大量probe overhead，且此方法可以在大型規模的網路下保持正常運作。 實驗結果顯示，我們提出的機制可以大幅改善70%的Direct Path，47%的Direct Path的RTT可以降低超過一半。而overhead與目前最精準的Relay Path Selection方法比較，在一開始僅有其11%，而隨著session數目的增加，overhead比例會越來越低。 總結以上，這個機制找尋Relay Path 非常有效率，它可以利用較少測量成本，找到接近最佳化的路徑，並且具有擴張性。以後，如果與其他路徑品質評比方法結合，還可以尋找到符合各種不同傳輸要求的路徑。

The demand of end-to-end communications is growing rapidly on Internet. The performance of end-to-end communications highly depends on the underlying Internet architecture. However, current Internet architecture, which has been used for several decades, architecture, was originally designed for data transmission purpose, instead end-to-end communications. Therefore, end-to-end communications do not perform well on Internet. Previously researches show that end-to-end communications can adopt the relay node, which is another node used to forward traffic between two ends, to greatly enhance the performance. But these researches introduce extremely huge probe overhead in the process of locating relays, or can not handle the large scale network well. This paper designs a Transit AS-Based Relay Path Selection Mechanism (TAB) for relay selection. The underlying idea of TAB is to use the characteristic of Internet architecture; that is, most of the ASs in Internet AS topology are Tasnit ASs. By applying TAB, we can reduce the probe overhead significantly in the Relay Path Selection process. Experiment results show that the proposed schemes can improve 70%of Direct Path, and 47% of Direct Path can reduced RTT by half. Furthermore, the overhead caused by TAB is only 11% of the one in the current most precise Relay Path Selection method, and the ratio of overhead decreases as the session number increases. Therefore, TAB is an effective Relay Path Selection method that can find nearly optimal path with lower probe overhead. Furthermore, TAB can adopt other metrics of path quality and find paths that fulfill various quality requirements.