基於多重向量場及頻寬估測之網路傳輸控制協定設計與分析

Abstract

針對TCP 之壅塞控制，本論文在此提出了三種不同的架構。前兩者分別稱為“Vegas-like TCP Type I”及“Vegas-like TCP Type II”，這兩種均是基於向量場的概念下而設計的，並且繼承了TCP-Vegas的特性，也就是利用封包傳輸的來回時間長度來作為網路壅塞的依據。Type I改變了演算法中向量場所能切換的動態數目，而Type II雖然保持著和TCP-Vegas相同的向量場動態切換數，但每一個動態形式均與Vegas有所不同。第三種則稱做“Bandwidth-based TCP”，與大部分TCP演算法不同的是，這裡利用了頻寬的估測來作為網路壅塞的依據，藉以控制window size的調變量。此演算法嘗試去估計window size的平衡點，進而讓其能在一個來回時間後達到該值。 在本論文中，一開始先對於TCP及AQM做了一個簡單的概述，而後針對幾個著名的演算法來做介紹，像是TCP-Vegas或是AQM-REM。接下來的章節則陸續說明了上述的三種機制，除此之外，透過相位圖以及線性化的分析，該機制的穩定度也被分別地探討。最後，藉由模擬的結果可知，不管是在對於網路頻寬的適應性，還是在同質網路或異質網路環境之下的公平性，本論文所提出的架構均展現了優於TCP-Vegas的效果。

This thesis presents three types of frameworks for TCP congestion control. The first two protocols, called “Vegas-like TCP Type I” and “Vegas-like TCP Type II”, respectively, are designed based on the vector field method and inherit the idea from a TCP-Vegas protocol that use RTT (round-trip-time) as a congestion measure. Type I gives a change in the number of switching dynamics of the vector field while Type II keeps the same number of switching dynamics of the vector field as Vegas but with a different style in each of them. The third one, called “Bandwidth-based TCP”, differs from most TCP algorithms by using the bandwidth estimation as the congestion measure to control the window size increment. It tries to predict the equilibrium point of window size then making the congestion window approach this point in a round-trip-time. In this thesis, an overview of TCP and AQM is given, and some well-known algorithms such as TCP-Vegas and AQM-REM are introduced subsequently. Then, three types of mechanisms are proposed in succession. In addition to giving the ideas, the stability of these mechanisms is also investigated via phase portrait method and linearization. Finally, through the simulations, the proposed scheme is shown to have a better performance over TCP-Vegas in adaptability and fairness both under a homogeneous and a heterogeneous environment.