摘要
随着互联网的日益发展，在互联网上的应用越来越丰富，这要求互联网能够提供更高的传输质量。如何改善互联网技术以满足新的业务需求，给广大研究人员提出了新的难题。网络虚拟化的提出使得这一难题得到解决，它为研究互联网新技术的改革和互联网新的基础架构提供了便利条件。
网络虚拟化就是在同一个物理环境中生成多个虚拟网络，虚拟网络之间逻辑上是完全隔离的，研究人员可以在虚拟网络上进行各种互联网新技术新协议的研究而不影响现有物理网络的运行。目前网络虚拟化的研究还处于理论研究阶段，其研究重点主要放在虚拟节点映射和虚拟链路映射的问题上。对于节点映射算法，本文提出以CPU和内存为共同参考量的节点映射算法，可以有效的避免内存成为限制节点映射瓶颈的情况。对于虚拟链路映射算法，本文提出改进的虚拟链路分割算法，它可以在一条物理链路上映射不能满足的情况下，将虚拟链路映射到多条物理链路上以满足映射的需求，这将提高物理资源的利用率，减少节点的二次映射问题。实验表明，通过将本文提出的节点映射算法和链路映射算法应用到网络虚拟化实验管理平台中，能够有效地均衡网络负载，得到良好的映射效果。
In the past half century, the development of information technology, especially theadvances in computer and internet technology has greatly changed the way people work andlive. The increasingly diverse network applications and network technologies to the existinginternet system has brought new challenges, such as flexibility, scalability and manageability,making the existing network architecture need to continuously change to adapt to the newnetwork technologies and application requirements. However, due to all the existing internetservice provider (ISP) can not be agreed upon a common goal and strategy, the innovation ofthe existing network architecture can only be very slow. At the same time, any newtechnology for the architecture for large-scale adjustment is difficult to test and deploy in thecurrent internet.
The virtualized network, making the problem solved. It allows multiple heterogeneousvirtual subnets in the virtual co-exist with a physical environment, rather than affecting thenormal operation of the existing network system. The nature of network virtualization is in apublic physical network, through the isolating mechanism relatively, relatively independent torun multiple virtual network to select the physical resources for the allocation and scheduling,making the physical network resources to maximize the use, reducing the network operatingcosts, improving the quality of network service. Network virtualization brings theseconvenient but also brings new challenges for network resource management.
The current status of the network virtualization at home and abroad are introduced in thispaper at first, highlighting several important international researching network virtualizationprojects, including projects such as GENI, as OpenFlow. Secondly, the key technologies ofnetwork virtualization are introduced, including the design of virtual nodes and virtual links,the virtual network mapping, the identity of the identification and addressing in the virtualnetwork, the management and scheduling for the virtual network resource, and so on. In thisarticle, we improve the mapping algorithm on the virtual network of the virtual nodes andvirtual links, and this algorithm is used in the experimental platform for networkvirtualization.
In the current, at home and abroad, studying the virtual node mapping algorithms most only considers the CPU utilization as the only resource, ignoring the impact of the memory.But in the actual mapping process, the memory size for generating virtual nodes has a greatinfluence. In this paper, for the virtual node mapping, we design a virtual node mappingalgorithm for the minimum busy. The so-called minimum busy refers to select the physicalnodes with more remaining CPU and memory firstly. Not just refer to the CPU as the onlyreference value. When there has enough memory, the CPU is as the main reference value.When the memory is not sufficient, the memory constraints limit the virtual node mapping.The value of busy degree can be the total CPU and memory resources compared to theremainder of the CPU and the remaining memory size. The smaller value descries the lowerbusy degree. This node mapping algorithm can effectively balance the physical network nodepressure to reduce the possibility of a bottleneck node, improving the physical network load,making the network resources used in balance.
For the virtual link mapping algorithm, this article describes a path splitting virtual linkmapping algorithm. The so-called path splitting mapping algorithm is when the virtual linkcorresponding the virtual node can not meet the mapping needs, we can consider this virtuallink is mapped on two or more physical links for meeting the virtual link requirements. Thisalgorithm largely avoided the current greedy mapping algorithm can not meet the linkmapping needs. When a physical node corresponding to a virtual node on the physical linkcan not meet the constraints of the virtual link, in accordance with the ordinary greedymapping algorithm, it needs to be virtual node re-mapping, to meet the constraints of thevirtual link. By using the path splitting link mapping algorithm, we can calculate thebandwidth of multiple physical links on this physical node, able to meet the constraints of thevirtual link and virtual nodes do not need to re-map. Therefore, this link mapping algorithmcan effectively reduce the probability of virtual node quadratic remapping, improve thephysical net work resources utilization, make the network link load balance and ensure thereliability of the link.
In the current, the research network virtualization technology is in the theoretical stage athome. There is no experiment platform for network virtualization. The fourth chapterdescribes the network virtualization experiment management platform; the platform is aprototype system of the experimental platform of network virtualization. Through thisplatform, users can submit virtual network request. The platform accepts the user’s request,according to the virtual node mapping algorithm and the virtual link mapping algorithm, mapsthe virtual nodes and virtual links in the request of the virtual network to the physical network.And the platform sends the configuration files to the physical nodes; the physical nodes receive the files, generating virtual nodes according to the configuration files, and configurethe virtual network to complete the virtual network mapping process. The users can login thevirtual network through the platform to configure their own network experiments for thevirtualization research.
The platform uses the minimum peak-degree virtual node mapping algorithm andpath-splitting virtual link mapping algorithm mapping the virtual requests. After repeatedrequests, the two mapping algorithms can be effective virtual network mapping.
引文
[1] N. Feamster, L. Gao, J. Rexford. How to lease the internet in your spare time[C].SIGCOMM Computer Communication Review,2007,37(1):61-64.
[2] J. Turner, D. Taylor. Diversifying the internet[C]. In Proc. of the IEEE GlobalTelecommunications Conference,2005.
[3] T. Anderson, L. Peterson, S. Shenker, J. Turner. Overcoming the Internet impassethrough virtualization[C]. IEEE Computer,2005,38(4):34–41.
[4]郭占军. GPRS VPN实现方式及其安全保障[J].科技资讯,2006,29:3.
[5] J. Carapinha, J. Jimenez. Network Virtualization–a View from the bottem [C]. in Proc.of the Workshop on Virtualized infrastructure systems and architectures,2009:73-80.
[6] D. Andersen, H. Balakrishnan, F. Kaashoek, R. Morris. Resilient overlay networks[C].SIGOPS Operating Systems Review,2001,35(5):131-145.
[7] Y. Chu, S. Rao, S. Seshan, H. Zhang. Enabling conferencing applications on the internetusing an overlay muilticast architecture [C]. SIGCOMM Computer CommunicationReview,2001,31(4):55-67.
[8] L. Subramanian, I. Stoica, H. Balakrishnan, R. Katz. OverQoS: An overlay basedarchitecture for enhancing internet QoS[C]. In Proc. of the1st Symposium on NetworkedSystems Design and Implementation (NSDI), March2004:71-84.
[9] UCLP项目[EB/OL].[2012-2-9]. http://uclp.uwaterloo.ca/.
[10]J. Recio, E. Grasa, S. Figuerola, G. Junyent. Evolution of the user controlled lightpathprovisioning system[C]. In Proc. of7th International Conference on Transparent OpticalNetworks,2005,1:263-266.
[11]B. Nandy, D. Bennett, I. Ahmad, S. Majumdar, B. St.Arnaud. User controlled lightpathmanagement system based on a service oriented architecture [EB/OL].http://www.solananetworks.com/UCLP/files/UCLPv2-SOA.pdf,2006.
[12]A. Sundararaj, P. Dinda. Towards virtual networks for virtual machine grid computing[C].In Proceedings of the3rd USENIX Virtual Machine Research and TechnologySymposium (VM'04),2004.
[13]VNET项目[EB/OL].[2012-2-10]. http://virtuoso.cs.northwestern.edu/.
[14]X. Jiang, D. Xu. VIOLIN: Virtual internetworking on overlay infrastructure [R]. PurdueUniversity, Tech. Rep, TR-03-027,2003.
[15]P. Ruth, X. Jiang, D. Xu, S. Goasguen. Virtual distributed environments in a sharedinfrastructure [M]. Computer,2005,38(5):63-69.
[16]A. Whitaker, M. Shaw, S. D. Gribble. Scale and performance in the denali isolationkernel[C]. In Proc. of the5th symposium on Operating Systems Design andImplementation (OSDI'02),2002:195-210.
[17]PlanetLab平台[EB/OL].[2012-2-11]. http://www.planet-lab.org/.
[18]VINI项目[EB/OL].[2012-2-11]. http://www.vini-veritas.net/.
[19]GENI项目[EB/OL].[2012-2-12]. http://www.geni.net/.
[20]S. F. Bush. Active virtual network management protocol[C]. In Proc. of the Workshop onParallel and Distributed Simulation,1999:182-192.
[21]OpenFlow项目[EB/OL].[2012-2-15]. http://www.openflow.org/.
[22]J. Lu, J. Turner. Efficient mapping of virtual networks onto a shared substrate[R].Technical Report WUCSE-2006-35, Washington University,2006.
[23]J. Fan, M. Ammar. Dynamic topology configuration in service overlay networks: A studyof reconfiguration policies[C]. In Proc. IEEE INFOCOM,2006.
[24]R. K. Ahuja, T. L. Magnanti, and J. B. Orlin. Network Flows:Theory, Algorithms, andApplications[M]. Prentice Hall,1993.
[25]E. Kohler, R. Morris, B. Chen, J. Jannotti, M. F. Kaashoek. The Click modular router[C].ACM Transactions on Computer Systems,2000,18(3):263-297.
[26]XORP软件[EB/OL].[2012-2-16]. http://www.xorp.org/.
[27]L. Mathy, N. Egi, M. Hoerdt, A. Greenhalgh, M. Handley.(Some) implementation issuesfor virtual routers[C]. In Workshop on Management of Network Virtualization,2007.
[28]J. Fan, M. Ammar. Dynamic topology configuration in service overlay networks-astudy of reconfiguration policies[C]. In Proc. of IEEE INFOCOM,2006.
[29]Y. Zhu, M. Ammar. Algorithms for assigning substrate network resources to virtualnetwork components[C]. In Proc. of IEEE INFOCOM,2006.
[30]M. Yu, Y. Yi, J. Rexford, M. Chiang. Rethinking virtual network embedding: Substratesupport for path splitting and migration[C]. ACM SIGCOMM Computer CommunicationReview, April2008,38(2):17-29.
[31]I. Houidi, W. Louati, D. Zeghlache. A distributed virtual network mapping algorithm[C].In Proc. of IEEE ICC,2008:5634–5640.
[32]N. M. M. K. Chowdhury, M. R. Rahman, R. Boutaba. Virtual Network Embedding withCoordinated Node and Link Mapping[C]. In Proc. of IEEE INFOCOM,2009:783-791.
[33]W. Szeto, Y. Iraqi, R. Boutaba. A multi-commodity flow based approach to virtualnetwork resource allocation[C]. In Proc. of the IEEE Global TelecommunicationsConference (GLOBE-COM'03),2003:3004-3008.
[34]C. Marquezan, J. Nobre, L. Granville, G. Nunzi, M. Brunner, D. Dudkowski. DistributedReallocation Scheme for Virtual Network Resources[C]. ICC’09,2009:1–5.
[35]Struts项目[EB/OL].[2012-2-20]. http://struts.apache.org/.
[36]Spring项目[EB/OL].[2012-2-20]. http://www.springsource.org/.
[37]Hibernate项目[EB/OL].[2012-2-20]. http://www.hibernate.org/.
[38]Cacti项目[EB/OL].[2012-2-20]. http://www.cacti.net/.
[39]Xen项目[EB/OL].[2012-2-20]. http://xen.org/.
[40]Y. Wang, E. Keller, B. Biskeborn, J. Merwe, J. Rexford. Virtual routers on the move:Live router migration as a network-management primitive[C]. In Proc. of the ACMSIGCOMM'08,2008:231-242.
[41]Dorigo M, Di Caro. The ant colony optimization meta-heuristic: New ideas inoptimization [M]. New York: McGraw-Hill,1999:11-32.
[42]Emulab项目[EB/OL].[2012-2-21]. http://www.emulab.net/.
