資料中心中軟體定義網路之負載平衡及路由機制

雲端計算的普及化推動著資料中心的發展，雲端應用日趨多元化，加上資料中心佈署範圍擴大，以至於資料中心必須處理多樣化且大量的工作負載，如何有效的管控資料中心資源且降低人力成本成為雲端計算所要面臨的問題，而軟體定義網路技術的出現為資料中心帶來嶄新的突破。針對雲端計算所面臨的挑戰，許多研究提出了多種網路架構，但現今大多資料中心使用三層架構的Multi-rooted Trees來提升靈活性和快速佈署。本研究中提出，我們提出了一個基於軟體定義網路的大型資料中心的網路架構，透過對每台Switch進行Topology-Aware的定址來自動化路由，不僅降低人為疏失也降低FIB的數量，此外集中式的管理不僅能夠進行全域性的路由管理也能夠避免廣播風暴和降低頻寬的浪費，封包的路徑選擇上搭配了On-Line Routing的方法計算最佳路徑，並且搭配多路徑路由來分散網路流量於網路拓樸中。我們使用Mininet模擬工具進行On-Line Routing和ECMP的比較，實驗結果顯示On-Line Routing比ECMP高出了60%的效能。

關鍵字

軟體定義網路；路由機制；負載平衡；資料中心

並列摘要

The Popularization of cloud computing facilitates the development of data centers. The applications of Cloud Computing have become more diverse and the scope of cloud computing is getting wider and wider, so data centers must deal with diverse set of workloads. The biggest challenge for cloud computing is how to effectively allocate resource in data centers to reduce operating costs. The emergence of software-defined networking technology creates a new breakthrough network technology for data centers. Some research works have adopted specially designed network topologies to cope with the challenges faced by cloud computing, such as Fat tree, but most data centers still adopt the multi-rooted-tree topology as the underlying infrastructure topology due to its flexibility and rapid development of high speed switches. In this research, we proposed the data center network architecture based on the Software-Defined Networks (SDN). By assigning topology-aware addresses to each switch, auto routing not only reduces human errors but also decreases the size of FIB. Besides, the centralized management can avoided broadcast storm and increase bandwidth utilization by global routing decisions. In term of path selection strategy, we adopted On-Line Routing (OLR) method to calculate the optimal path. We also studied the effect of distributing network traffic via multipath routing. We compared the performance of OLR with that of ECMP via running simulations using the mininet emulator. Our numerical results show that the throughput of OLR increases 60% as compared to that of ECMP.

並列關鍵字

Software Defined Network ； Routing Mechanism ； Load Balancing ； Data Center

參考文獻

[4] S. Agarwal, M. Kodialam, and T. Lakshman, “Traffic engineering in Software Defined Networks,” in INFOCOM, 2013, pp. 2211–2219.

[5] M. Al-Fares, A. Loukissas, and A. Vahdat, “A Scalable, Commodity Data Center Network Architecture,” ACM SIGCOMM Computer Communication Rev., vol. 38, no. 4, 2008, pp. 63–74.

[8] M. Al-Fares, S. Radhakrishnan, B. Raghavan, N. Huang, and A. Vahdat, “Hedera: Dynamic Flow Scheduling for Data Center Networks.” in NSDI, vol. 10, 2010, pp. 19.

[11] A. R. Curtis, W. Kim, and P. Yalagandula, “Mahout: Low-Overhead Datacenter Traffic Management using End-Host-Based Elephant Detection,” in INFOCOM, 2011 Proc. IEEE, 2011, pp. 1629–1637.

[16] J. Zhang, F. Ren, T. Huang, L. Tang, and Y. Liu, “Congestion-aware adaptive forwarding in datacenter networks,” Computer Communications, vol. 62, 2015, pp. 34–460.

國際替代計量

資料中心中軟體定義網路之負載平衡及路由機制

未授權

主題瀏覽