跳到主要內容

臺灣博碩士論文加值系統

(44.213.60.33) 您好!臺灣時間:2024/07/20 04:36
字體大小: 字級放大   字級縮小   預設字形  
回查詢結果 :::

詳目顯示

我願授權國圖
: 
twitterline
研究生:王晟凱
研究生(外文):Wang, Sheng-Kai
論文名稱:應用於OPTUNS資料中心網路之先進軟體定義網路隧道控制系統
論文名稱(外文):Advanced SDN-based Tunnel Control for OPTUNS Data Center Networks
指導教授:田伯隆
指導教授(外文):Tien, Po-Lung
口試委員:楊啟瑞施汝霖
口試委員(外文):Maria YuangShih, Ju-Lin
口試日期:2020-9-1
學位類別:碩士
校院名稱:國立交通大學
系所名稱:電機工程學系
學門:工程學門
學類:電資工程學類
論文種類:學術論文
論文出版年:2020
畢業學年度:109
語文別:英文
論文頁數:51
中文關鍵詞:軟體定義網路邊緣資料中心網路流量控制OpenFlow
外文關鍵詞:SDNEdge Data CenterTraffic ControlOpenFlow
相關次數:
  • 被引用被引用:0
  • 點閱點閱:197
  • 評分評分:
  • 下載下載:0
  • 收藏至我的研究室書目清單書目收藏:1
第五代行動通訊系統(5G)為許多物聯網應用(IoT)的基礎,其中包含像是車聯網、智慧工廠、虛擬實境、擴增實境等等,以上應用針對低延遲和高頻寬等方面要求極高,但受限於裝置本身的計算能力,資料通常會透過網際網路被傳送至雲端進行處理,處理完成後的資訊再被傳送回裝置,但大部分的雲端資料中心距離使用者或是物聯網裝置是有段距離的,此距離會造成端到端(end-to-end)的延遲增加,將不足以應付以上5G應用低延遲的要求,而邊緣資料中心(Edge Data Center)就是為了解決此問題而生,藉由將資料中心布建在更靠近使用者或物聯網裝置的地方,資料透過網路傳送至資料中心的延遲可有效降低。

另一方面,將資料中心布建在更靠近使用者端後並不能完全解決問題,原因在於靠近使用者端的網路流量(Network Traffic)將會比在雲端資料中心所面臨到的更加不穩定且突發,若是沿用現有成熟的Spine/Leaf電交換機資料中心架構(Electrical Spine/Leaf Data Center, ESLDC),並利用等價多路徑路由(Equal-cost Multi-path Routing, ECMP)進行路由決策時,會有無法應付突發流量而造成延遲增加的風險,而這就是智慧定義光隧道網路系統(Optical Intelligence-defined Tunnel Network System, OPTUNS)被設計出來的契機,其具備諸多優點,包含提供巨量頻寬、高可擴充性、高可靠性、低延遲、低耗能等等。

而基於OPTUNS之上,同樣由我們實驗室所研發的軟體定義網路(Software-defined Networking)系統除了用來控制OPTUNS外,更是實現超低延遲、即時動態光波長配置和路由最佳化等目標的重要推手。本篇論文的研究對象,即是針對已開發的系統進行優化與改進,重要改進包含兩項,分別為(一)變更Top-of-Rack(TOR)交換機中的flow entry格式,用以針對現實情況中,一個機櫃中的伺服器IP通常為不連續且隨機的情況,和(二)一個全新的流量分配機制,使所有光隧道的負載能被盡可能的平均分配,除了能降低延遲外,也可以減少光隧道超載(Overload)的情況發生。

實驗結果顯示,新的flow entry設計不僅支援了隨機份配的IP地址分配,同時間盡可能的減少總共需要在交換機中下的flow entry 數量,此優勢不但能使系統支援的IP地址數量更多,更能使我們所需監測的flow entry數量大量減少,進而讓監測整個OPTUNS的程序更加快速且即時,而新的流量分配機制除了盡可能的使不同的光通道間負載更平均之外,其演算法對系統造成的負擔和相對應所需要的執行時間也被控制在低於2毫秒,達到對於光隧道overload/underload情況做最快速的處理,並且在系統的資料吞吐量(throughput)和延遲(latency)達到有效的平衡。
The fifth generation mobile communication system (5G) is the basis of numerous IoT-related applications, including Internet of Vehicle (IoV), smart factory, virtual reality(VR), augmented reality (AR), etc. All of the above-mentioned applications have strict requirements regarding latency and bandwidth. However, limited by the computer power they can possess, data generated by these devices often needs to be sent to a cloud data center over the Internet, processed, then be sent back to the devices to be utilized. This process can inherit a high end-to-end latency, which will not satisfy the strict latency requirement of 5G applications, and this is where edge data center comes from. By constructing smaller scale but still powerful data center closer to the user or IoT devices at the edge, the latency caused by sending data back and forth could be greatly suppressed.

However, once we bring the data center from cloud to edge, the traffic patterns it is facing will be very different from those it is facing at the cloud, that is to say the traffic becomes more unstable and more bursty, which causes problem for the current state-of-the-art data center to handle. The current state-of-the-art data center utilizes electrical switches to form a spine-leaf structure, thus we call it electrical spine-leaf data center(ESLDC), and it depends on equal-cost multi-path routing (ECMP) to make routing decisions. While these two technologies have gone through many tests and are proved to be mature and error-free, it can incur higher latency when dealing with traffics which are very bursty and with high locality. On the other hand, Optical Intelligence-defined Tunnel Network System (OPTUNS) is designed to solve the problem. Not only does it guarantee ultra-low latency, but also provide huge bandwidth, high scalability, high reliability, low power consumption, etc. Namely, it has the full capability to replace the widely-used electrical spline-leaf data center network architecture.

On top of OPTUNS, we develop a software-defined networking (SDN) controller, which not only controls OPTUNS, but also performs as the key that makes ultra-low latency, real-time dynamic wavelength distribution and optimized routing possible. Based on the already-developed controller designed by the past researchers, the goal of this thesis is to make it more optimized and suitable for real-world applications. There are two major modifications. The first one is that we modify the flow entry format defined in the top-of-rack(TOR) switches so that it can adapt to the situation where servers’ IPs under each rack are not continuous. The second one is a new overload/underload handling algorithm based on a load balancing criterion, which not only balances the loads of optical tunnels, but also reduces the traffic latency.


Experiment results show that, the new flow entry design not only supports random server IP address, but also suppresses the number of required flow entries by a factor the number of IP addresses under a rack. This enables the system to accommodate more IP addresses and also reduces the number of flow entries that we need to monitor, making the monitoring process of OPTUNS more real-time. Regarding the overload/underload handling algorithm, test results shows that the algorithm is extremely light and fast and it only takes less than two milliseconds to execute, achieving real-time handling of the overloaded/underloaded tunnel.
摘要. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . v
Abstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . vii
Acknowledgement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ix
Table of Contents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . x
List of Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xii
1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.1 Background and Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Thesis Organization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
2 OPTUNS Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
2.1 Hardware Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
2.1.1 Network Architecture . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
2.1.2 Optical Tunnels . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.2 Software Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
2.2.1 Main Modules . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
2.2.2 Tunnel Scheduler . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.2.3 Bandwidth Usage Monitor . . . . . . . . . . . . . . . . . . . . . . . . 14
3 Rack-to-Server based Flow Entry Design . . . . . . . . . . . . . . . . . . . . . . . 16
3.1 Flow Entry Definition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
3.2 Consecutive-IP Flow Entry Design . . . . . . . . . . . . . . . . . . . . . . . . 18
3.3 Server-to-Server Based Flow Entry Design . . . . . . . . . . . . . . . . . . . . 20
3.4 Rack-to-Server based Flow Entry Design . . . . . . . . . . . . . . . . . . . . 23
4 Load Balancing based Overload/Underload Handling Algorithm . . . . . . . . . 27
4.1 Overload Handling Algorithm . . . . . . . . . . . . . . . . . . . . . . . . . . 27
4.2 Underload Handling Algorithm . . . . . . . . . . . . . . . . . . . . . . . . . . 33
4.3 Heuristic Load Balancing Function . . . . . . . . . . . . . . . . . . . . . . . . 35
5 Performance Evaluation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
5.1 Monitoring Interval . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
5.2 Flow Modification Execution Time . . . . . . . . . . . . . . . . . . . . . . . . 43
5.3 Overload/Underload Handling Execution Time . . . . . . . . . . . . . . . . . 45
6 Conclusion and Future Works . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
6.1 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
6.2 Future Works . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
[1] Meng-Ru Tsai, “Design and Implementation of 360°Virtual Reality Live Streaming Based
on OPTUNS,” Master’s thesis, National Chiao-Tung University, 2019.
[2] M. Yuang, P. Tien, W. Ruan, T. Lin, S. Wen, P. Tseng, C. Lin, C. Chen, C. Chen, Y. Luo,
M. Tsai, and S. Zhong, “Optuns: Optical intra-data center network architecture and prototype
testbed for a 5g edge cloud [invited],” IEEE/OSA Journal of Optical Communications
and Networking, vol. 12, no. 1, pp. A28–A37, 2020.
[3] OpenDaylight Project, Boron SR3 ed., Pica8, March 2017, Available: https://
docs.opendaylight.org/en/latest/downloads.html.
[4] Chun-Ting Chen, “Design and Implementation of SDN Traffic-Monitoring based Optical-
Tunnel Control for Optical Edge Datacenter Network,” Master’s thesis, National Chiao-
Tung University, 2019.
[5] M. Yuang, P.-L. Tien, H.-Y. Chen, W. Ruan, T.-K. Hsu, S. Zhong, J. Zhu, Y. Chen, and
J. Chen, “Opmdc: Architecture design and implementation of a new optical pyramid data
center network,” Journal of Lightwave Technology, vol. 33, pp. 1–1, 05 2015.
[6] OpenFlow Switch Specification, 1.3.2 ed., Open Networking Foundation, April 2013,
Available: https:// www.opennetworking.org/ wp-content/ uploads/ 2014/10/ openflowspec-
v1.3.2.pdf.
[7] PICOS Documentation, 2.11.24 ed., OpenDaylight Foundation, November
2019, Available: https:// docs.pica8.com/ display/ PicOS21124sp/
Flow+Scalability+per+Broadcom+Chipset.
[8] HPE ProLiant DL380 Gen10 Server Specification, Delta Products Corporation, August
2020, Available: https:// h20195.www2.hpe.com/ v2/ GetDocument.aspx? docname=
a00008180enw.
[9] Delta AG7648 Specification, Hewlett Packard Enterprise, Hune 2020, Available: https://
agema.deltaww.com/product-info.php?id=29.
[10] Manpage of TCPDUMP, TCPDUMP, August 2020, Available: https://www.tcpdump.org/
manpages/tcpdump.1.htmllbAI.
連結至畢業學校之論文網頁點我開啟連結
註: 此連結為研究生畢業學校所提供,不一定有電子全文可供下載,若連結有誤,請點選上方之〝勘誤回報〞功能,我們會盡快修正,謝謝!
QRCODE
 
 
 
 
 
                                                                                                                                                                                                                                                                                                                                                                                                               
第一頁 上一頁 下一頁 最後一頁 top
無相關期刊