隨著各種云服務(wù)的部署增加,今天的數(shù)據(jù)中心網(wǎng)絡(luò)變得越來越大。隨著網(wǎng)絡(luò)規(guī)模的不斷擴(kuò)大,細(xì)粒度網(wǎng)絡(luò)監(jiān)控成為了網(wǎng)絡(luò)可靠性和閉環(huán)流量控制的先決條件。然而,由于控制平面和數(shù)據(jù)平面之間的連續(xù)交互以及有限的CPU能力,這種監(jiān)視機(jī)制是粗粒度的,并且導(dǎo)致在具有高密度數(shù)據(jù)中心網(wǎng)絡(luò)的現(xiàn)代網(wǎng)絡(luò)中無法很好地隨劇烈變化的通信動(dòng)態(tài)擴(kuò)展。為了增加可擴(kuò)展性,P4語言聯(lián)盟（P4.org）提出了帶內(nèi)網(wǎng)絡(luò)遙測(cè)（INT）機(jī)制,其提供實(shí)時(shí)數(shù)據(jù)平面監(jiān)視網(wǎng)絡(luò)。在本文中,我們提出了帶內(nèi)網(wǎng)絡(luò)遙測(cè)的概念,并提出使用路由路徑生成策略來解決問題。在WINT系統(tǒng)中,我們可以精確地控制每個(gè)探測(cè)路徑,從而可以為非重疊INT路徑生成一個(gè)多樣的INT路徑規(guī)劃算法來覆蓋整個(gè)網(wǎng)絡(luò)拓?fù)鋱D并確保路徑數(shù)最小,并對(duì)算法運(yùn)行是的復(fù)雜度進(jìn)行詳盡分析。在這里,我們首先提出一種基于深度優(yōu)先搜索（DFS）的簡單算法,該算法簡單且省時(shí)。其次,我們提出了基于歐拉路徑的算法來最優(yōu)地生成具有最小路徑數(shù)的非重疊INT路徑。為了實(shí)現(xiàn)更好的負(fù)載平衡,經(jīng)典數(shù)據(jù)中心中網(wǎng)絡(luò)拓?fù)涫羌嫒莸?并且存在多種多路徑拓?fù)淠Ｊ�。因此INNT路徑非常適合在數(shù)據(jù)中心網(wǎng)絡(luò)中部署。另外,仿真結(jié)果表明歐拉算法的執(zhí)行時(shí)間... 

【文章來源】：北京郵電大學(xué)北京市 211工程院校 教育部直屬院校

【文章頁數(shù)】：60 頁

【學(xué)位級(jí)別】：碩士

【文章目錄】：
ABSTRACT
摘要
Chapter 1: Introduction
    1.1 Research Background
    1.2 Traditional networking
    1.3 Software-Defined Network(SDN)
        1.3.1 Control plane:
        1.3.2 Data plane:
    1.4 Active network
    1.5 OpenFlow
    1.6 Data plane programmability
    1.7 Contribution
    1.8 Summary
    1.9 Thesis Organization
Chapter 2: Software-defined networking for computer networking
    2.1 Overview of Software-Define Networking
        2.1.1 Intelligence and speed
        2.1.2 Easy network management:
        2.1.3 Multi-tenancy:
        2.1.4 Network Security
    2.2 Traditional vs Software Define Networking
    2.3 Software-Defined Networking Architecture
        2.3.1 Data layer/Infrastructure layer
        2.3.2 Control layer
        2.3.3 Application Layer
    2.4 Feature of SDN
    2.5 Open Flow
    2.6 Programmable Data plane
        2.6.1 Protocol Oblivious Forwarding
        2.6.2 P4 (Programming Protocol-Independent Packet Processing)
    2.7 Network wide telemetry in Data center network
        2.7.1 Tree based topologies
            Basic tree
            FatTree network topology
            Leaf-Spine network topology
    Summary
Chapter 3: Algorithm Design
    3.1 System Background:
    3.2 Challenging issue:
        3.2.1 Telemetry overhead
        3.2.2 Synchronized multi-path monitoring
    3.3 Problem statement:
    3.4 The Proposed Algorithms
    3.5 Depth-First Search algorithm
    3.6 Euler Trail based algorithm
        3.6.1 Algorithm Description
        3.6.2 Case study
        3.6.3 Run-time complexity analysis:
        3.6.4 Heuristic method for balanced path generation
    Summary
Chapter 4: Evaluation
    2.1 Experimental setting
    2.2 Experimental Results
        2.2.1 Number of generated INT path:
        2.2.2 Balanced and Unbalanced Path Generation:
        2.2.3 Telemetry Overhead:
        2.2.4 Execution time of path planning algorithms
    2.3 Network-wide telemetry in Data Center (DC) Network
Chapter 5: Conclusion
ACKNOWLEDGEMENT
REFERENCES



本文編號(hào)：3327805