一個處理不當高頻寬串流之改良式RED方法

Abstract

RED機制是一種著名的動態緩衝區管理(active queue management)的機制，可以用來控制網路上的擁塞狀況，而且RED機制已經被推薦必須廣泛地應用在網際網路的路由器上。然而在網路發生擁塞的時期，RED機制並不能夠保護那些有著良好行為的網路資料流(和TCP協定相容的網路資料流)免於受到有些使用高頻寬且有著錯誤行為的網路資料流(包括那些沒有使用擁塞控制機制的網路資料流，以及那些有著不好的TCP協定實作的網路流量)的傷害。我們提出了一種新的機制來改善RED機制的能力，使得改善後的RED機制能夠處理那些使用高頻寬且有著錯誤行為的網路資料流。由實驗模擬後所得到的結果中可以觀察到我們所提出的機制成功地分辨出那些使用高頻寬且有著錯誤行為的網路資料流，並且加以懲罰。此外，從模擬實驗的結果中，我們發現所提出的方法對於所有共同競爭的網路資料流可以達到趨近公平性的效果。

The Random Early Detection (RED) mechanism is a well-known active queue management mechanism that was related to congestion control. And the RED mechanism has been recommended to be deployed in the Internet router. However, the RED mechanism cannot protect well-behaved flows (i.e. TCP-compatible flows) against the high-bandwidth misbehaving flows, these include flows that do not use congestion avoidance mechanism and flows with poorly implemented TCP protocol, during periods of congestion. We proposed a new mechanism to improve RED’s functionality to deal with the high-bandwidth misbehaving flows. The simulation results show that our proposed mechanism can successfully identify and penalize the high-bandwidth misbehaving flows. Besides, from those simulation results, we could find that our proposed mechanism can achieve approximate fairness among all competing flows.