跳到主要內容

臺灣博碩士論文加值系統

(44.222.104.206) 您好!臺灣時間:2024/05/28 00:26
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:盧咨亦
研究生(外文):LU,Zih-Yi
論文名稱:AMT-TCP:應用在資料中心TCP之可調適標記閥值
論文名稱(外文):AMT-TCP: Adaptive marking threshold for Data Center TCP
指導教授:詹益禎
指導教授(外文):Chan,Yi-Cheng
口試委員:詹益禎何承遠李春良
口試委員(外文):Chan,Yi-ChengHo, Cheng-YuanLee,Chun-Liang
口試日期:2021-08-24
學位類別:碩士
校院名稱:國立彰化師範大學
系所名稱:資訊工程學系
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:2021
畢業學年度:109
語文別:中文
論文頁數:94
中文關鍵詞:資料中心網路壅塞控制
外文關鍵詞:data center networkcongestion controlTCP IncastDCTCPNS-2
相關次數:
  • 被引用被引用:0
  • 點閱點閱:95
  • 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:0
中文摘要
雲端服務近年來蓬勃的發展,在電競、高畫質影音、物聯網、AI、社群平台…等服務持續的拓展下,資料中心已成為雲端服務不可或缺的基礎建設。雲端服務所帶來的便利性顯而易見,加上近年來許多企業陸續地建立起屬於自己的資料中心,加速資料中心網路的發展,引起許多學者想對其做更深入的探討及研究。資料中心網路主流的傳輸層協定為TCP,但資料中心網路內部常見的多對一傳輸方式是傳統TCP新的挑戰,當來源端的數量一旦過多且同時發送到通一個接收端時,在連接鏈路中的緩衝區不夠大的情況下會造成大量的封包遺失,更嚴重會導致網路效能的崩潰,這個現象被稱為 TCP Incast[6]。

在論文中,我們針對資料中心網路TCP提出利用中間節點標記輔助和修改發送端壅塞控制機制的方案來改善TCP Incast問題以提高網路的使用效率,此方法稱為 Adaptive marking threshold for Data Center TCP (AMT-TCP)。AMT-TCP利用頻寬延遲乘積(BandwidthDelay-Products, BDP)和交換器的緩衝區大小(switch buffer size)作為設置網路中間節點壅塞標記閥值的依據,並據此積極調整壅塞窗口大小以更敏銳的方式取代原先的TCP壅塞控制機制。

在NS-2的模擬實驗下,我們將AMT-TCP的效能與CTTCP和DCTCP做比較,AMT-TCP在多組的實驗參數下皆有較佳的表現。整體而言AMT-TCP在資料中心網路中有著更彈性和更有效率的行為,並且能夠有效的解決封包遺失事件避免TCP Incast的現象發生,在TCP剛啟動時就可以較快的速度達到滿載的網路效能,NS-2模擬器的實驗結果證實了AMT-TCP的有效性。

關鍵字 : 資料中心網路、壅塞控制、TCP Incast、DCTCP、NS-2
The vigorous development of cloud services has gradually matured and stabilized. With the continuous expansion of e-sports, high-definition audiovisual, Internet of Things, AI, social platforms, etc., data center networks have become an indispensable infrastructure for cloud services, The convenience brought by cloud services is obvious. In addition, in recent years, many companies have successively established their own data center networks, accelerating the development of data center networks, and even causing people to want to do more in-depth discussions and research on them. The transmission protocol of the data center network is currently the mainstream TCP/IP. The many-to-one transmission method inside the data center network is somewhat different from the traditional TCP. When the number of sources is too large and they are sent to the same receiver at the same time, the connection chain If the buffer in the road is not large enough, a large number of packets will be lost, and even more serious will lead to the breakdown of network performance. We call this phenomenon TCP Incast.
In this paper, we will improve the TCP Incast problem and increase the utilization rate of the network by using the mark-assisted solution of the intermediate node for the data center network to improve the TCP Incast problem and increase the network utilization rate through and modify the sender congestion control mechanism. Adaptive marking threshold for Data Center TCP (AMT-TCP). This method uses the bandwidth-delay product (BDP) and the switch buffer size as the basis for adjusting the network threshold value, and adjusts the speed of the sending window in a faster way Replacing the original congestion control, AMT-TCP can use bandwidth faster and effectively during data center network transmission, avoiding packet loss events.
After many simulations of NS-2 experiments, we compare the performance of AMT-TCP with CTTCP and DCTCP. AMT-TCP has better performance under the condition of multiple sets of experimental parameters. On the whole, it is known from the experimental results that AMT-TCP has a more flexible and efficient performance in the data center network, and can effectively solve the packet loss event to avoid the TCP Incast phenomenon. The fastest speed achieves the best network performance.
Keywords: data center network, congestion control, TCP Incast, DCTCP, NS-2
目錄
中文摘要…………………………………………………………………………………......VI
英文摘要……………………………………………………………………………………VII
誌謝………………………………………………………………………………………..... IX
目錄……………………………………………………………………………………….......X
圖目錄 ………………………………………………………………………………..…..XII
表目錄…………………………………………………………………………………...…XVI
第一章 緒論…………………………………………………………………………………1
第二章 文獻探討……………………………………………………………………………3
2.1 TCP Incast…………………………………………………………………………..5
2.2 Tahoe…..…………………………………………………....………………………6
2.3 Reno……………………………………………………………………………...…7
2.4 New Reno…………………………………………………………………………...7
2.5 Sack …………………………………………………………………………………7
2.6 DCTCP…………………………………………………………………………... …8
2.7 CTTCP...………………………………………………………………………….....9
第三章 AMT-TCP……..……………………………….………………………………….13
3.1研究動機………………………………………………………………….……….......13
3.2研究方法………………………………………………………………….……….......14
3.2.1以頻寬、往返時間及交換器緩衝區大小作為壅塞控制門檻值設置的依據…......15
3.2.2以更積極的方式調整發送窗口取代傳統緩啟動及壅塞避免機制………….........17
第四章 實驗結果…………………………………………………………………………..21
4.1.AMT-TCP、CT-TCP與DCTCP之比較…………………………………….............22
4.1.1伺服器數量為2, 交換器緩衝區大小為256KB, 網路往返時間為100µs……….22
4.1.2伺服器數量為2, 交換器緩衝區大小為256KB, 網路往返時間為200µs……….27
4.1.3伺服器數量為8, 交換器緩衝區大小為256KB, 網路往返時間為100µs……….32
4.1.4伺服器數量為8, 交換器緩衝區大小為256KB, 網路往返時間為200µs……….36
4.1.5伺服器數量為32, 交換器緩衝區大小為256KB, 網路往返時間為100µs………41
4.1.6伺服器數量為32, 交換器緩衝區大小為256KB, 網路往返時間為200µs………46
4.1.7伺服器數量為2, 交換器緩衝區大小為4MB, 網路往返時間為100µs………….51
4.1.8伺服器數量為2, 交換器緩衝區大小為4MB, 網路往返時間為200µs…………56
4.1.9伺服器數量為8, 交換器緩衝區大小為4MB, 網路往返時間為100µs………….61
4.1.10伺服器數量為8, 交換器緩衝區大小為4MB, 網路往返時間為200µs…………66
4.1.11伺服器數量為32, 交換器緩衝區大小為4MB, 網路往返時間為100µs……….71
4.1.12伺服器數量為32, 交換器緩衝區大小為4MB, 網路往返時間為200µs………76
4.1.13伺服器數量為1, RTT為100µs及緩衝區256KB………………………………..82
4.1.14伺服器數量為1, RTT為100µs及緩衝區4MB…………………………………...83
4.2.三個版本SRU時間…………………………………………………………...………85
4.2.1.伺服器為2、緩衝區為256KB ……………………………………….……………85
4.2.2.伺服器為2、緩衝區為4MB ……………………………………….……………...86
4.2.3.伺服器為32、緩衝區為256KB ……………………………………….…………...87
4.2.4.伺服器為32、緩衝區為4MB ……………………………………….……………..89
第五章 結論……………………………………………………………………………......91
參考文獻………………………………………………………………………………….....92
參考文獻
[1] M. Alizadehz, A. Greenberg, D. A. Maltz, J. Padhye, P. Patel, B. Prabhakar, S. Sengupta, M. Sridharan, “Data Center TCP (DCTCP),” SIGCOMM’10, pp.63-74, August 30-September 3, 2010.
[2] M. Al-Fares, A. Loukissas, A. Vahdat, “A scalable, commodity data center network architecture”, in proceedings of the ACM SIGCOMM’08, pp.63-74, August 2008.
[3] T. Benson, A. Anand, A. Akella, M. Zhang, ”Understanding Data Center Traffic Characteristics,” in Proceedings of ACM WREN 2009.
[4] V. Jacobson, “Congestion avoidance and control,” in Proc. of ACM SIGCOMM, pp.314-329, August 1988.
[5] The Network Simulator- ns-2. Available at http://www.isi.edu/nsnam/ns/.
[6] A. Phanishayee, E. Krevat, V. Vasudevan, D. G. Andersen, G. R. Ganger, G.A. Gibson, and S. Seshan “Measurement and analysis of TCP throughput collapse in cluster-based storage systems,” in Proceedings of the 6th USENIX Conference on File and Storage Technologies (FAST’08), pp.1-14, February 2008.
[7] G. Appenzeller, I. Keslassy, and N. McKeown. “Sizing Router Buffers. ” In ACM Sigcomm, 2004.
[8] Y. Chen, R. Griffith, J. Liu, A. Joseph, R.H. Katz, “Understanding TCP Incast Throughput Collapse in Datacenter Networks,” Workshop on Research in Enterprise Networks (WREN’09), pp.73-82, August 2009.
[9] A. Kulseitova, Ang-Tan Fong, “A survey of energy-efficient techniques in cloud data centers,” ICT for Smart Society, pp.1-12, June 2013.
[10] H. Wu, Z. Feng, C. Guo, Y. Zhang, “ICTCP: Incast Congestion Control for TCP in Data Center Networks,” IEEE/ACM Transactions on Networking, vol.21, no.2, pp. 345-358, April 2013.
[11] R.P. Tahiliani, M.P. Tahiliani and K.C. Sekaran “TCP Congestion Control in Data Center Networks,” Handbook on Data Centers, pp. 485-505, March 2015.
[12] Alizadeh, M., Javanmard, A., & Prabhakar, B. (2011). Analysis of DCTCP: stability, convergence, and fairness. ACM SIGMETRICS Performance Evaluation Review, 39(1), 73-84.
[13] Chen, Y., Griffith, R., Liu, J., Katz, R. H., & Joseph, A. D. (2009, August). Understanding TCP incast throughput collapse in datacenter networks. In Proceedings of the 1st ACM workshop on Research on enterprise networking (pp. 73-82).
[14] Das, T., & Sivalingam, K. M. (2013, January). TCP improvements for data center networks. In 2013 Fifth International Conference on Communication Systems and Networks (COMSNETS) (pp. 1-10). IEEE.
[15] Chen, W., Cheng, P., Ren, F., Shu, R., & Lin, C. (2013, July). Ease the queue oscillation: Analysis and enhancement of dctcp. In 2013 IEEE 33rd International Conference on Distributed Computing Systems (pp. 450-459). IEEE.
[16] Nichols, K., & Jacobson, V. (2012). Controlling queue delay. Communications of the ACM, 55(7), 42-50.
[17] Gettys, J. (2011). Bufferbloat: Dark buffers in the internet. IEEE Internet Computing, 15(3), 96-96.
[18] Sikdar, B., Kalyanaraman, S., & Vastola, K. S. (2003). Analytic models for the latency and steady-state throughput of TCP Tahoe, Reno, and SACK. IEEE/ACM Transactions On Networking, 11(6), 959-971.
[19] Wang, S., Zhang, J., Huang, T., Pan, T., Liu, J., & Liu, Y. (2020). A-ECN Minimizing Queue Length for Datacenter Networks. IEEE Access, 8, 49100-49111.
[20] Wang, S., Zhang, J., Huang, T., Pan, T., Liu, J., & Liu, Y. (2017, October). Adaptively adjusting ECN marking thresholds for datacenter networks. In 2017 IEEE 25th International Conference on Network Protocols (ICNP) (pp. 1-2). IEEE.
[21] CT-TCP:用於資料中心網路的門檻值控制TCP, July 2018.
電子全文 電子全文(網際網路公開日期:20250902)
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