本文關(guān)鍵詞：Fat-Tree網(wǎng)絡(luò)中的多播與節(jié)能技術(shù)研究　出處：《西安電子科技大學(xué)》2014年碩士論文　論文類型：學(xué)位論文

  更多相關(guān)文章： 胖樹 輸入緩存 交叉開關(guān)矩陣 低扇出分割多播調(diào)度 節(jié)能

【摘要】：隨著網(wǎng)絡(luò)的迅猛發(fā)展,尤其是IPTV、視頻會(huì)議、遠(yuǎn)程呈現(xiàn)等帶寬密集型服務(wù)的流行,多播交換技術(shù)因其能以資源節(jié)約的方式分發(fā)業(yè)務(wù)而變得越來(lái)越重要。然而,如果一個(gè)分組去往多個(gè)目的地,資源競(jìng)爭(zhēng)發(fā)生的可能性會(huì)更大。隨著當(dāng)前網(wǎng)絡(luò)流量的日益增加,對(duì)高速、大容量、綠色節(jié)能的多播交換技術(shù)的研究日趨重要。本文主要研究基于輸入緩存Crossbar的多播調(diào)度算法以及在此基礎(chǔ)上應(yīng)用于Fat-Tree網(wǎng)絡(luò)的節(jié)能算法。1.設(shè)計(jì)了一種低扇出分割多播調(diào)度算法。該算法是種迭代算法,應(yīng)用改進(jìn)的隊(duì)列結(jié)構(gòu)及動(dòng)態(tài)權(quán)值計(jì)算方法,使得輸出端口能盡量選擇同一個(gè)信元,從而減少扇出分割次數(shù),可以分為信元預(yù)處理、迭代匹配、分組傳輸、隊(duì)頭更新四個(gè)階段,其中迭代匹配過(guò)程是該算法的核心。改進(jìn)了一種隊(duì)列組織結(jié)構(gòu),在傳統(tǒng)的虛擬輸出排隊(duì)(VOQ)的基礎(chǔ)上,增加了多播狀態(tài)字隊(duì)列、多播暫存VOQ以及獨(dú)立隊(duì)列,可以進(jìn)一步緩解隊(duì)頭阻塞。設(shè)計(jì)了一種動(dòng)態(tài)權(quán)值計(jì)算算法。在調(diào)度多播信元時(shí),與現(xiàn)有調(diào)度算法每次迭代時(shí)多播信元的權(quán)值都保持固定不同,新算法在每輪迭代中根據(jù)多播信元的扇出分割情況動(dòng)態(tài)地為信元計(jì)算權(quán)值,以確保為扇出分割小的信元提供更多優(yōu)先輸出機(jī)會(huì)。設(shè)計(jì)了一種雙向移位優(yōu)先級(jí)設(shè)置方法。在計(jì)算動(dòng)態(tài)權(quán)值時(shí)需要引用輸入端口當(dāng)前時(shí)隙的優(yōu)先級(jí),傳統(tǒng)的優(yōu)先級(jí)計(jì)算方法存在一定的不公平性,對(duì)此,本文中設(shè)計(jì)了一種雙向移位優(yōu)先級(jí)設(shè)置方法,該方法具有良好的公平性,并對(duì)其進(jìn)行了證明。低扇出分割多播調(diào)度算法將上述各個(gè)算法協(xié)調(diào)統(tǒng)一為一個(gè)系統(tǒng),并保證信元間無(wú)亂序等問(wèn)題。該算法在傳統(tǒng)的隨機(jī)流量模型及固定流量模型下測(cè)試結(jié)果良好。2.設(shè)計(jì)了一種基于Fat-Tree網(wǎng)絡(luò)的節(jié)能算法。針對(duì)Fat-Tree網(wǎng)絡(luò)的拓?fù)浣Y(jié)構(gòu)設(shè)計(jì)了一種簡(jiǎn)單但有效的路由算法,并設(shè)計(jì)了一種節(jié)能算法。調(diào)度算法采用低扇出分割多播調(diào)度算法,并對(duì)其進(jìn)行了少量的修改以適應(yīng)路由算法和節(jié)能算法。在Bernoulli及Bursty業(yè)務(wù)源下,分別采用不同比例的單多播流量模型進(jìn)行仿真測(cè)試,測(cè)試結(jié)果良好。
[Abstract]:With the rapid development of the network, especially IPTV, video conferencing, remote rendering such as bandwidth intensive services popular multicast switching technology because of the way to save resources and distribution business becomes more and more important. However, if a packet to multiple destinations, the possibility of resource competition will be greater with the current network traffic is increasing, the high speed, large capacity, multicast green energy exchange technology is becoming more and more important. This thesis focuses on the input memory Crossbar multicast scheduling algorithm and.1. algorithm based on Fat-Tree network is applied to design a low fanout multicast scheduling algorithm. The segmentation the algorithm is an iterative algorithm, calculation method and application of improved queue structure dynamic weights, the output port to select the same cell, thereby reducing the fanout splitting times, can be divided into cell The pretreatment, iterative matching, packet transmission, update the team head in four stages, the iterative matching process is the core of the algorithm. An improved queue structure in traditional virtual output queuing (VOQ) based on the increase of multicast queue status word, multicast and temporary VOQ independent cohort, can to further ease the head of line blocking. Design a dynamic weight calculation algorithm. The scheduling multicast cell, and the existing scheduling algorithm in each iteration when the multicast cell weights are kept fixed, the new algorithm in each iteration according to the multicast cell fan segmentation dynamically calculate the weights for the cell, to ensure as the fanout splitting small cell to provide more output opportunities. Priority to design a bidirectional shift priority setting method. Need to reference the input port of the current slot priority in the calculation of dynamic weight calculation method, the priority of the traditional There is no fairness, some of this, this paper designs a bidirectional shift priority setting method, this method has good fairness, and has carried on the proof. Low fanout splitting multicast scheduling algorithm the algorithm will be unified into a system, and to ensure that no problem between cell disorder. The algorithm test results good.2. design an energy saving algorithm based on Fat-Tree network. In the traditional model of random traffic flow model and fixed for Fat-Tree network topology is a simple but efficient routing algorithm is designed, and an energy saving algorithm design. Scheduling algorithm with low fanout splitting multicast scheduling algorithm. And the little modification to adapt to routing algorithm and energy saving algorithm. In the Bernoulli and Bursty business sources, respectively with different ratios of single multicast traffic simulation model test, test The result is good.

【學(xué)位授予單位】：西安電子科技大學(xué)
【學(xué)位級(jí)別】：碩士
【學(xué)位授予年份】：2014
【分類號(hào)】：TN915.0

【相似文獻(xiàn)】

相關(guān)碩士學(xué)位論文 前1條

1 祝劍鋒;Fat-Tree網(wǎng)絡(luò)中的多播與節(jié)能技術(shù)研究[D];西安電子科技大學(xué);2014年

，

本文編號(hào)：1380041