本文關(guān)鍵詞：基于傳輸片段部分重疊的多路徑傳輸技術(shù)　出處：《北京交通大學(xué)》2017年碩士論文　論文類型：學(xué)位論文

  更多相關(guān)文章： 多路徑傳輸 在線視頻 重疊

【摘要】：在線觀看高清視頻已經(jīng)成為普遍的網(wǎng)絡(luò)行為,并且占用了大量網(wǎng)絡(luò)帶寬。繼續(xù)靠增加網(wǎng)絡(luò)帶寬提升用戶觀看體驗(yàn)的空間已經(jīng)不大。多路徑傳輸技術(shù),即,通過客戶端的多個(gè)網(wǎng)絡(luò)接口實(shí)現(xiàn)數(shù)據(jù)在多條路徑上共同傳輸,能夠打破這種瓶頸�，F(xiàn)有多種多路徑傳輸技術(shù)。一種典型的方法是,采取在多條路徑上同時(shí)傳輸完全相同數(shù)據(jù);以保證數(shù)據(jù)及時(shí)到達(dá),但是,這種方法會(huì)造成網(wǎng)絡(luò)資源浪費(fèi)。另一種典型的方法是,根據(jù)預(yù)測(cè)的帶寬為每條路徑分配相應(yīng)大小的數(shù)據(jù)傳輸任務(wù),以實(shí)現(xiàn)數(shù)據(jù)同時(shí)到達(dá)的目標(biāo);但是,由于路徑帶寬符合高斯變化,在一次數(shù)據(jù)塊傳輸過程中,越晚被傳輸?shù)臄?shù)據(jù)(數(shù)據(jù)塊尾部),越容易發(fā)生傳輸超時(shí);一旦尾部數(shù)據(jù)無(wú)法及時(shí)到達(dá),客戶端播放質(zhì)量就會(huì)受到影響。本文利用路徑帶寬的高斯變化,提出了一種新的多路徑傳輸思想。按照這種思想,客戶端下載數(shù)據(jù)的過程被分割為不同的下載周期;每個(gè)周期,客戶端需要從服務(wù)器下載一定大小的數(shù)據(jù)塊以供播放。由于路徑帶寬符合高斯變化,這將導(dǎo)致每個(gè)周期內(nèi),一條路徑上較晚傳輸?shù)臄?shù)據(jù)(尾部數(shù)據(jù))傳輸成功的概率較低;因此,對(duì)尾部數(shù)據(jù)進(jìn)行重復(fù)傳輸并調(diào)節(jié)重復(fù)傳輸?shù)臄?shù)據(jù)量,可以提高尾部數(shù)據(jù)成功傳輸?shù)母怕?從而最大化一個(gè)周期內(nèi)數(shù)據(jù)成功傳輸?shù)母怕�。同時(shí),因?yàn)橹粚?duì)尾部數(shù)據(jù)進(jìn)行重復(fù)傳輸,所以引起的開銷又不會(huì)太大。本文的主要工作和貢獻(xiàn)如下:(1)本文建立了基于部分?jǐn)?shù)據(jù)重疊的多路徑傳輸技術(shù)的傳輸模型。在這一模型中,我們清晰地描述了本方法是如何最大化一個(gè)周期內(nèi)數(shù)據(jù)成功傳輸?shù)母怕省?2)本文提出了基于部分?jǐn)?shù)據(jù)重疊的多路徑傳輸技術(shù),PCT(Partial Overlapping Chunk Based Mutiple-Path Transmission)。本方法在每個(gè)周期內(nèi),為了以最大概率下載到所需的數(shù)據(jù)塊,客戶端會(huì)根據(jù)每條路徑的傳輸能力(路徑帶寬的期望和方差),安排不同的數(shù)據(jù)傳輸任務(wù):各路徑所請(qǐng)求獲取的數(shù)據(jù)具有相同的終止位置以及不同起始位置,而數(shù)據(jù)的起始位置由各路徑帶寬的高斯變化共同決定。(3)本文通過理論分析以及一系列的仿真實(shí)驗(yàn)證明了 PCT。與各路徑傳輸完全不同數(shù)據(jù)的方法相比,基于模型的仿真分析顯示PCT以最高14.4%的冗余率為代價(jià),可以最多提高19.3%的傳輸成功率;更進(jìn)一步,基于真實(shí)網(wǎng)絡(luò)的仿真分析則顯示PCT在路徑總帶寬較小時(shí),對(duì)傳輸成功率的優(yōu)化更明顯。
[Abstract]:Viewing HD video online has become a common network behavior and occupies a large amount of network bandwidth. There is no longer much space to improve the user's viewing experience by increasing network bandwidth. Multipath transmission technology, that is. It is possible to break this bottleneck by means of multiple network interfaces of the client. A typical method is to transfer data over multiple paths. The same data is transmitted at the same time on multiple paths; In order to ensure the timely arrival of data, however, this method will lead to waste of network resources. Another typical method is to assign data transmission tasks of corresponding size for each path according to the predicted bandwidth. In order to achieve the goal that the data arrive at the same time; However, because the path bandwidth is in line with Gao Si's change, the later the data is transmitted in a data block transmission process, the more likely the transmission timeout will occur. Once the tail data can not arrive in time, the quality of client playback will be affected. This paper proposes a new multipath transmission idea based on Gao Si's change of path bandwidth. The process of downloading data by client is divided into different downloading cycles. Each cycle, the client needs to download a certain size of data blocks from the server for playback. Because the path bandwidth conforms to Gao Si changes, this will lead to each cycle. The probability of successful transmission of late data (tail data) on a path is lower; Therefore, repeat transmission of tail data and adjust the amount of data transferred repeatedly can improve the probability of successful transmission of tail data, thereby maximizing the probability of successful transmission of data in a period. At the same time. Because only the tail data is transferred repeatedly. The main work and contribution of this paper are as follows: 1) this paper establishes a transmission model of multipath transmission based on partial data overlap. We clearly describe how this method maximizes the probability of successful data transmission within a period. PCT(Partial Overlapping Chunk Based Mutiple-Path Transmission). This method is within each cycle. In order to download to the desired data block with the maximum probability, the client will base on the transmission capacity of each path (the expectation and variance of the path bandwidth). Schedule different data transfer tasks: the data requested by each path has the same stop position and different starting position. The starting position of the data is determined by the variation of the bandwidth of each path. Through theoretical analysis and a series of simulation experiments, this paper proves that PCT. is compared with the method of transmitting completely different data from each path. The simulation analysis based on the model shows that PCT can increase the transmission success rate of 19.3% at the cost of up to 14.4% redundancy rate. Furthermore, the simulation analysis based on real network shows that PCT is more effective in optimizing the transmission success rate when the total path bandwidth is small.
【學(xué)位授予單位】：北京交通大學(xué)
【學(xué)位級(jí)別】：碩士
【學(xué)位授予年份】：2017
【分類號(hào)】：TN919.8

【參考文獻(xiàn)】

相關(guān)期刊論文 前10條

1 楊洋;楊家海;溫皓森;王會(huì);;基于SDN流量測(cè)量的數(shù)據(jù)中心多路徑傳輸研究[J];華中科技大學(xué)學(xué)報(bào)(自然科學(xué)版);2016年11期

2 薛開平;陳珂;倪丹;張泓;洪佩琳;;基于MPTCP的多路徑傳輸優(yōu)化技術(shù)綜述[J];計(jì)算機(jī)研究與發(fā)展;2016年11期

3 黃偉建;楊海龍;;基于Dijkstra算法分布式JobTracker節(jié)點(diǎn)模型通信方式的優(yōu)化[J];河南師范大學(xué)學(xué)報(bào)(自然科學(xué)版);2016年03期

4 陳鳴;胡慧;劉波;邢長(zhǎng)友;許博;;一種基于OpenFlow的多路徑傳輸機(jī)制[J];電子與信息學(xué)報(bào);2016年05期

5 杜白;李紅艷;;多優(yōu)先級(jí)的異構(gòu)網(wǎng)絡(luò)多路徑并行視頻傳輸算法[J];信號(hào)處理;2016年04期

6 楊洋;楊家海;秦董洪;;數(shù)據(jù)中心網(wǎng)絡(luò)多路徑路由算法[J];清華大學(xué)學(xué)報(bào)(自然科學(xué)版);2016年03期

7 YAN Jinyao;ZHANG Hailong;SHUAI Qianjun;LIU Bo;GUO Xiao;;HiQoS:An SDN-Based Multipath QoS Solution[J];中國(guó)通信;2015年05期

8 趙禮峰;嚴(yán)子恒;;基于增廣鏈修復(fù)的最大流求解算法[J];計(jì)算機(jī)應(yīng)用;2015年05期

9 郭湘軍;周堯;吳書新;;MATLAB在電子信息專業(yè)課程教學(xué)中的主要應(yīng)用[J];電子技術(shù)與軟件工程;2014年06期

10 曹宇;徐明偉;;一種按需分配的多路徑傳輸分組調(diào)度算法[J];軟件學(xué)報(bào);2012年07期

相關(guān)碩士學(xué)位論文 前2條

1 戴承石;全國(guó)排球聯(lián)賽技術(shù)統(tǒng)計(jì)平臺(tái)現(xiàn)狀調(diào)查與分析[D];蘇州大學(xué);2015年

2 陳均華;多路徑傳輸中亂序與負(fù)載均衡研究[D];浙江大學(xué);2010年

，

本文編號(hào)：1416617