【摘要】：作為一種非常重要的傳輸層協(xié)議以及網(wǎng)絡(luò)中端到端擁塞控制的主要組成部分,TCP在各種網(wǎng)絡(luò)環(huán)境中的性能一直以來都受到廣泛關(guān)注。TCP的性能主要是由其擁塞控制算法決定的。與此同時如今具有大的時延帶寬積的高速網(wǎng)絡(luò)正在隨著技術(shù)的進(jìn)步而逐漸出現(xiàn)并被應(yīng)用于實際的生產(chǎn)和生活環(huán)境中。近些年來,針對這樣的網(wǎng)絡(luò),有很多的擁塞控制算法被提出來用以改進(jìn)TCP的性能。本文研究了高速網(wǎng)絡(luò)上TCP擁塞控制算法,主要包括以下的三部分工作：提出一種新的TCP擁塞控制算法、利用NS-2仿真平臺驗證算法性能以及在Linux系統(tǒng)上實現(xiàn)并在實際的網(wǎng)絡(luò)中測試。 現(xiàn)如今的網(wǎng)絡(luò)中有很多不同類型的高速TCP擁塞控制算法,可以大致分成基于丟包、基于延遲以及前兩者的混合這三種類型。但是其中基于丟包的這些高速的TCP協(xié)議存在潛在的對于已經(jīng)在網(wǎng)絡(luò)中廣泛應(yīng)用的Reno的不公平性,這會嚴(yán)重的損害在同一條瓶頸鏈路上Reno的吞吐量。另一方面,純粹的基于延遲的流在和基于丟包的流的競爭中沒有優(yōu)勢。在本文中,我們提出了一種基于控制論的新的混合的TCP擁塞控制算法：Hybrid High Speed-TCP(HHS-TCP)。它是一種基于丟包和基于延遲思想的協(xié)同方法。HHS-TCP的發(fā)送速度由兩種不同的機(jī)制共同決定。新算法能夠在當(dāng)鏈路的帶寬沒有得到充分利用時快速增長發(fā)送速度,當(dāng)瓶頸鏈路排隊延遲增大則減小發(fā)送速率。HHS-TCP提供了良好的帶寬利用率的擴(kuò)展性,同時也能提供較好友好性。 為了驗證我們提出的新算法的性能,作者使用了NS-2仿真平臺做了較為全面的測試。通過對隨機(jī)丟包鏈路單個流的吞吐量以及分別在Drop-tail和RED兩種不同的路由策略下多個流的吞吐量的測量,可以看出HHS-TCP的帶寬利用率更高；同時通過比較不同類型的高速TCP協(xié)議與Reno之間在同一條瓶頸鏈路共同競爭時的吞吐量的比例,看出混合型的HHS-TCP能夠取得較好的公平性。此外為了能夠在實際的網(wǎng)絡(luò)中測試新算法的性能以及應(yīng)用到實際的網(wǎng)絡(luò)業(yè)務(wù)中,在分析了Linux中TCP協(xié)議棧結(jié)構(gòu)的基礎(chǔ)上,我們將新算法在Linux系統(tǒng)中以模塊化的形式實現(xiàn),方便加載和刪除。在實際的局域網(wǎng)中使用新的模塊進(jìn)行文件傳輸?shù)臏y量,結(jié)果顯示新算法能夠比之前的協(xié)議提高10%-15%左右。 本文的主要貢獻(xiàn)和創(chuàng)新有：本文通過研究網(wǎng)絡(luò)中的擁塞問題以及TCP中的擁塞控制算法,分析了擁塞產(chǎn)生的原因以及對擁塞控制算法的評價方法,提出了一種新的基于混合機(jī)制的擁塞控制算法,該算法能在高速網(wǎng)絡(luò)中獲取高的帶寬利用率的同時保持對傳統(tǒng)TCP協(xié)議的公平性。同時我們將新算法通過模塊化的方式在Linux系統(tǒng)中加以實現(xiàn),實際測試顯示新算法具有一定的實用價值。
[Abstract]:As a very important transport layer protocol and the main component of end-to-end congestion control in the network, the performance of TCP in various network environments has been widely concerned. The performance of TCP is mainly determined by its congestion control algorithm. At the same time, the high speed network with large delay bandwidth product is emerging gradually with the development of technology and applied to the actual production and living environment. In recent years, many congestion control algorithms have been proposed to improve the performance of TCP. In this paper, the TCP congestion control algorithm in high-speed networks is studied, which includes the following three parts: a new TCP congestion control algorithm is proposed. The NS-2 simulation platform is used to verify the performance of the algorithm, and the algorithm is implemented on the Linux system and tested in the actual network. Nowadays, there are many different types of high-speed TCP congestion control algorithms in networks, which can be divided into three types: loss of packet, delay and mixing of the first two. However, these high-speed TCP protocols based on packet loss have potential unfairness to the Reno which has been widely used in the network, which will seriously damage the throughput of Reno on the same bottleneck link. On the other hand, pure delay-based flows have no advantage over packet-based flows. In this paper, we propose a new hybrid TCP congestion control algorithm named: hybrid High Speed-TCP (HHS-TCP) based on cybernetics. It is a cooperative method based on packet loss and delay. The transmission speed of HHS-TCP is determined by two different mechanisms. The new algorithm can rapidly increase the transmission speed when the bandwidth of the link is not fully utilized, and reduce the transmission rate. HHS-TCP provides a good scalability of bandwidth utilization and good friendliness when the bottleneck link queue delay increases. In order to verify the performance of the proposed new algorithm, the author has made a more comprehensive test using the NS-2 simulation platform. By measuring the throughput of a single stream in a random packet loss link and the throughput of multiple streams under two different routing strategies of Drop-tail and RED, we can see that the bandwidth utilization of HHS-TCP is higher. At the same time, by comparing the ratio of throughput between different types of high-speed TCP protocols and Reno in the same bottleneck link competition, we can see that hybrid HHS-TCP can achieve better fairness. In addition, in order to test the performance of the new algorithm in the actual network and to apply it to the actual network service, on the basis of analyzing the TCP protocol stack structure in Linux, we implement the new algorithm in the Linux system in the form of modularization. Easy to load and delete. In the actual LAN, the new module is used to measure the file transfer. The results show that the new algorithm can increase the accuracy by about 10% to 15% compared with the previous protocol. The main contributions and innovations of this paper are as follows: by studying the congestion problem in the network and the congestion control algorithm in TCP, the causes of congestion and the evaluation method of congestion control algorithm are analyzed. A new congestion control algorithm based on hybrid mechanism is proposed, which can achieve high bandwidth utilization in high-speed networks while maintaining fairness to traditional TCP protocols. At the same time, we implement the new algorithm in Linux system by modularization. The practical test shows that the new algorithm has some practical value.
【學(xué)位授予單位】：北京郵電大學(xué)
【學(xué)位級別】：碩士
【學(xué)位授予年份】：2014
【分類號】：TP393.04

【參考文獻(xiàn)】

相關(guān)期刊論文 前2條

1 文宏;向德生;鄧小科;朱培棟;唐玉華;;FAST TCP協(xié)議研究及性能分析[J];計算機(jī)工程與應(yīng)用;2006年11期

2 于梅爾·哈克;;奧巴馬團(tuán)隊?wèi)?zhàn)斗規(guī)則[J];商界(評論);2009年02期

，

本文編號：2166752