端點對端點流量控制探測方式之效能分析

Abstract

解決網路壅塞的方法除了靠增加網路頻寬之外，就是要利用流量控制機制來避免網路壅塞狀況的產生；現有的機制大概可分為Window-Based與Rate-Based兩大類，Rate-Based流量控制機制使用探測的方法取得網路頻寬的資訊，以此作為流量控制的參考，因此利用測量方式取得網路的有效頻寬是個重要的課題。 現有取得網路頻寬資訊的方式大概有下列幾種：Throughput、Round Trip Time、Pathchar、Packet-Pair Probing。Packet-Pair Probing是一種應用於Rate-Based流量控制機制的探測網路頻寬方法，但是它的缺點就是很容易被網路上其他因素所干擾，造成所測量出的有效網路頻寬值並不準確，所以在EERCA中提出了一個Bounded Probing的方法來改進這個缺點。 在本篇論文中針對這兩種測量網路頻寬資訊的方式加以分析與比較。我們先以數學的方式分析出Packet-Pair Probing 以及Bounded Probing兩種方式所測量出有效網路頻寬的正確率，然後再以電腦模擬方式取得的實驗數據，將數學分析出的結果加以驗證。

End-to-End Flow Control is an efficient method to moderate the network congestion. The End-to-End Flow Control mechanisms can be categorized into two approaches. One is the Rate-Based Flow Control and the other is the Window Based Flow Control. In the former we can obtain the network bandwidth information through a probing mechanism, therefore how to measure the accurate network available bandwidth is very important. Current techniques used to obtain network bandwidth information are as follows: throughput, round trip time, pathchar, and packet-pair. In Rate-Based Flow Control, we can use the packet-pair scheme to probe the bandwidth of the network. But packet-pair cannot probe the available bandwidth precisely. In EERCA, a Bounded-Probing mechanism is proposed for reducing the probing error probability. In this thesis, we analyze the performance between the Bounded Probing scheme proposed in EERCA and Packet-Pair probing scheme. At first we analyze these two schemes mathematically, then we evaluate the analytical result by simulation.