基於軟體定義網路之多用戶大型資料中心設計

在本文中，我們提出了一個可靠且適用於大型資料中心的網路架構，由於這個設計混合了Ethernet與SDN交換機，可以有效降低建置成本。我們利用Ethernet交換機MAC learning與簡化狀態管理的好處，降低了環境設定的複雜度。另外，藉由在資料中心的核心網路採用SDN交換機，使得系統能夠動態更換路徑與快速故障修復。此網路架構在單一資料中心，能夠支援超過10萬台實體機與100萬台虛擬機同時在線，可容納超過64K個用戶，每個用戶可使用4096個虛擬網路，並且能夠自由指派IP位址與VLAN IDs。由於資料中心網路充斥著大量的廣播封包造成浪費頻寬，其中又以ARP封包為主，我們攔截了所有的ARP封包，由SDN controller負責回應所有的位址查找請求。為了增加系統的可靠性，使用了兩台SDN controller，透過in-band control channel，管理與控制網路中的任何裝置，並實作了隨插即用的功能，。我們為資料中心內的實體機啟用了MPTCP功能，提升TCP連線的吞吐量，並提出一個能夠避免使用較壅塞Link的congestion aware (CA)路徑選擇方法。我們在本文提到的網路架構中，實際進行模擬與實體測試，實驗結果表示，系統能夠處理非常大量的連線請求以及在進行Live Migration後，短時間內恢復被中斷的虛擬機連線。我們的實驗結果顯示，congestion aware路徑選擇方法搭配MPTCP與基於ECMP的MPTCP相比，無論是模擬還是實際測試，都獲得了較好的結果，因此虛擬機在大型資料中心啟用MPTCP可以獲得較多的網路吞吐量。

關鍵字

負載平衡；可擴展性；軟體定義網路；大型資料中心網路；虛擬機即時轉移；自動設定

並列摘要

Abstract— In this thesis, we propose an architecture for realizing a reliable large scale data center network. This network is a cost effective hybrid SDN-Ethernet network. We fully take advantage of the automatic learning capability of the Ethernet switches to simplify state management. By employing SDN switches in the core of the network, the system is able to perform route changes for supporting VM live migration and fast failure recovery. Using the proposed architecture, a single data center can support hundreds of thousand physical machines (PM) and own up to a million of virtual machines (VM). Besides, more than 64K tenants can share a single data center and each tenant can individually own up to 4K VLANs in their network. Each tenant is allowed to freely assign IP addresses and VLAN IDs to their VM. Unlike conventional IP network that employing packet broadcasting to handle ARP queries, in our system, ARP packets are considered to be control packets. They are processed in the system controller so as to mitigate the impact of ARP broadcast in a large data center To enhance reliability, dual SDN controllers are used to manage the network through in-band control channels. The controllers are able to detect network topology and computing servers automatically. It fully supports plug and play service to reduce the possible configuration errors introduced by human. We apply MPTCP to provide multiple subflows to enhance throughput of TCP connections. In this work, a novel congestion aware routing is propose for network load balancing. We have conducted experiments and simulations to evaluate the performance of the proposed network. The experimental results show that the system is able to handle heavy connection requests and resume connections after VM migrations within short time. Our simulation results also reveal that the proposed congestion aware routing outperforms ECMP. It improves the total MPTCP throughput significantly.

並列關鍵字

Load balanced ； Scalability ； Software Defined Network ； Large Data Center Network ； VM Live Migration ； Auto Configuration

參考文獻

[4] C. Guo, G. Lu, H. J. Wang, S. Yang, C. Kong, P. Sun, W. Wu, and Y. Zhang, “SecondNet: A Data Center Network Virtualization Architecture with Bandwidth Guarantees,” In ACM CoNEXT, 2010.

[7] M. Kodialam, T. V. Lakshman, and S. Sengupta. Efficient and Robust Routing of Highly Variable Traffic. In HotNets,2004..

[9] M. Chiesa, G. Kindler, and M. Schapira, “Traffic Engineering with ECMP:An Algorithmic Perspective,” In INFOCOM, 2014.

[13] N. Omnes, M. Bouillon, G. Fromentoux and O. Le Grand, “A programmable and virtualized network & IT infrastructure for the internet of things: How can NFV & SDN help for facing the upcoming challenges, ” In 18th International Conference on Intelligence in Next Generation Networks(ICIN), 2015.

[14] Li Li and Chou Wu, “Design and Describe REST API without Violating REST: A Petri Net Based Approach, ” In ICWS, 2011, 4-9 July 2011, Page(s) 508 – 515

國際替代計量

基於軟體定義網路之多用戶大型資料中心設計

未授權

主題瀏覽