透過封包改寫機制提升區域網路裝置效能之方案研究

Performance Improvement Schemes for Network Devices in Local Area Networks via Packet Rewriting

邱建樺(Chien-Hua Chiu)

指導教授 ： 雷欽隆

國立臺灣大學/電機資訊學院/電機工程學研究所/博士(2012年)

https://doi.org/10.6342/NTU.2012.00338

摘要

封包改寫對於在實體網路上建構虛擬網路而言是一種常用且有效的技術。在這篇論文中，我們以封包改寫技術為基礎，針對平行處理以及大規模乙太網路等研究領域提出了相關的研究方案。首先，針對於乙太網路環境下的網路裝置，我們提出了一個分散式的叢集式架構。透過封包改寫的方法，使得網路流量得以平行地分流到叢集中的各個網路裝置，進而提高整體的吞吐量。叢集中的網路裝置能夠分配其流量至其他裝置，使得系統具有負載平衡與系統容錯的能力。此架構更已成功導入不同的應用層如頻寬管理器、網路式入侵防禦系統等。第二、針對大規模乙太網路的研究，我們也利用封包改寫的方法，提出了一個新穎的網路裝置 – EtherAgent。它能夠大幅減少大規模乙太網路環境下的廣播封包，並且能夠向下相容於現有的乙太網路，而不用針對目前廣為使用的乙太網路修改其相關的網路協定及軟硬體。根據我們所做的實驗結果，針對我們以封包改寫為基礎所提出的網路效能改善方案，的確能夠有效地應用在網路裝置諸如頻寬管理器與網路式入侵服昱系統的叢集化。此外，同樣以封包改寫為技術基礎的EtherAgent，同樣也能應用在大規模乙太網路的環境下，有效地控制廣播封包的數量。

關鍵字

分散式叢集架構 ； 平行處理 ； 乙太網路 ； 負載平衡 ； 系統容錯 ； 大規模網路 ； 企業網路

並列摘要

Packet rewriting is a useful skill for creating virtual segments in the networks. In this dissertation, we propose several schemes based on packet rewriting in the field of parallel processing and the broadcast traffic reduction in large scale Ethernet. First, we propose a decentralized clustering scheme for network devices which operate in transparent mode. Devices in the cluster process the network traffic in parallel in a decentralized manner to scale the throughput. A device can also migrate its workload to others for the purpose of load balance or fault tolerance. The proposed clustering scheme is successfully applied to applications such as bandwidth controller and network intrusion prevention systems. Second, we propose a new device, EtherAgent, to reduce the broadcast traffic in the Ethernet. An EtherAgent is backward compatible so that it can be inserted into an existing Ethernet without any change on protocols, hardware, and software modifications. According to performance evaluation, the proposed schemes based on packet rewriting can be effectively used for building clusters of network devices such as bandwidth controller and network intrusion prevention systems. Moreover, the proposed network device, EtherAgent, can also significantly reduce broadcast traffic in large scale Ethernet.

並列關鍵字

Decentralized clustering scheme ； parallel processing ； Ethernet ； load balance ； fault tolerance ； large scale network ； enterprise network

參考文獻

[1] L. Aversa and A. Bestavros, “Load Balancing a Cluster of Web Servers: Using Distributed Packet Rewriting,” Proc. IEEE Int’l Conf. Performance, Computing, and Communications (IPCCC ’00), pp. 24-29, Feb. 2000.

[2] A. Bestavros, M. Crovella, J. Liu, and D. Martin, “Distributed Packet Rewriting and Its Application to Scalable Server Architectures,” Proc. IEEE Int’l Conf. Network Protocols (ICNP ’98), pp. 290-297, Oct. 1998.

[6] Z. Cao, Z. Wang, and E. Zegura, “Performance of Hashing-Based Schemes for Internet Load Balancing,” Proc. IEEE Infocom, pp. 323-341, 2000.

[7] V. Cardellini, E. Casalicchio, M. Colajanni, and P.S. Yu, “The State of the Art in Locally Distributed Web-Server Systems,” ACM Computing Surveys (CSUR), Vo. 34, No. 2, pp. 263-311, Jun. 2002.

[12] R. Droms. RFC 2131: Dynamic Host Configuration Protocol, 1997.