混合乙太網路與SDN之大型資料中心網路

Hybrid Ethernet and SDN Large Scale Data Center Network

陳毅融(I-Jung Chen)

指導教授 ： 李詩偉

國立中正大學/工學院/通訊工程學系/碩士(2015年)

摘要

在本論文中，我們提出了一個大型資料中心網路的架構。使用Ethernet 交換機置放在core層以及SDN 交換機置放在Aggregation層。充分的運用Ethernet 交換機的優點包含Layer 2 MAC 學習機制轉發封包、MSTP (Multiple Spanning Tree Protocol)可分配多路徑用以支援負載平衡以及斷路修復、SNMP (Simple Network Management Protocol)讓NMS(Network Management System)可以蒐集Ethernet交換機的相關資訊、價格低廉以及容易佈建支援隨插即用功能。藉由使用SDN交換機可程式化的彈性，我實作flooding 避免機制、ARP處理與封包更改、負載平衡路由、以及斷路修復機制。使用OpenStack建制管理大型資料中心的VM、PM，將環境實作在in-band架構下節省網路端口，以及實作NMS多項功能包含網路拓樸探詢、Spanning tree instance自動設定、鏈路負載收集以及斷路偵測。藉由混合Ethernet交換機及SDN交換機實做高效能低成本的大型資料中心網路。

關鍵字

軟體定義網路 ； 大型資料中心網路 ； 斷路修復 ； 負載平衡 ； 多生成樹協定 ； 簡單網路管理協定

並列摘要

In this work, we propose an architecture for a large scale data center network. This is a hybrid Ethernet-SDN network in which the Ethernet switches and the SDN switches are deployed in the core part and in the aggregation part of the network. We use commercial off-the-shelf (COTS) Ethernet switches to reduce CAPEX. The automatic MAC learning of the Ethernet switches are also helpful to simplify the state management and the implementation of plug-and-play function in the network. We use SDN switches to enable flexibility, programmability, and optimization of network resources, In this work, we focus on designing flooding free ARP processing, load balancing routing, and fast failure recovery. Experimental results show that our system can eliminate unnecessary flooding messages. Compared to random shortest path based routing, our congestion aware routing can provide fair bandwidth usages among different communication pairs. The network is featured by its fast failover capability. The average failure recovery time is less than 100 msec for any single link failure in the network.

並列關鍵字

Multiple Spanning Tree Protocol ； Software Defined Network ； Large scale data center ； Failure recovery ； Load balancing ； Simple Network Management Protocol

參考文獻

[2]T.Benson, A.Akella, A,Shaikh, S.Sahu” CloudNaaS: A Cloud Networking Platform for Enterprise Applications,” in Proc. ACM SOCC, 2011.

[3]C. Guo, G. Lu, H.J. Wang, S. Yang, C. Kong , P. Sun , W. Wu , Y. Zhang” SecondNet: A Data Center Network Virtualization Architecture with Bandwidth Guarantees,” in Proc. ACN SOCC, 2011.

[4]H. Cho, S. Kang, Y. Lee” Centralized ARP Proxy Server Over SDN Controller To Cut Down ARP Broadcast In Large-scale Data Center Networks,” in Proc. ICOIN 2015.

[5]V. G. Oliveira, M. Rogerio S., G. Curiel, D. Santos” Automatic Discovery of Physical Topology in Ethernet Networks,” in Proc. AINA, 2008.

[6]D. Santos, A. de Sousa, F. Alvelos, M. Dzida, M. Pi´oro and M. Z.˙ zon” Traffic Engineering of Multiple Spanning Tree Routing Networks: the Load Balancing Case,” in Proc. NGI , 2009.