过去20年中可编程网络的发展可以分为几个阶段？每个阶段的贡献是什么？
三个阶段。主动网络（从20世纪90年代中期到21世纪初），在网络中引入了可编程功能；控制和数据平面分离（从2001年到2007年左右），开发了控制和数据平面之间的开放接口；OpenFlow API和网络操作系统（2007年至2010年左右），代表了广泛采用开放接口的第一个实例，并开发了使控制数据平面分离可扩展且实用的方法。
Making computer networks more programmable en-ables innovation in network management and lowers thebarrier to deploying new services. In this section, wereview early work on programmable networks. We di-vide the history into three stages, as shown in Figure 1.Each stage has its own contributions to the history: (1) ac-tive networks (from the mid-1990s to the early 2000s),which introduced programmable functions in the network to enable greater to innovation; (2) control and data planeseparation (from around 2001 to 2007), which developedopen interfaces between the control and data planes; and(3) the OpenFlow API and network operating systems(from 2007 to around 2010), which represented the firstinstance of widespread adoption of an open interface anddeveloped ways to make control-data plane separationscalable and practical.
网络虚拟化与SDN的关系？
Net-work virtualization (an abstraction of the physical net-work in terms of a logical network) clearly does not re-quire SDN. Similarly, SDN (the separation of a logicallycentralized control plane from the underlying data plane)does not imply network virtualization. Interestingly, how-ever, a symbiosis between network virtualization and SDNhas emerged, which has begun to catalyze several new re-search areas. SDN and network virtualization relate inthree main ways:
•SDN as an enabling technology for network virtualiza-tion.Cloud computing brought network virtualization toprominence, because cloud providers need a way to al-low multiple customers (or “tenants”) to share the samenetwork infrastructure.Nicira’s Network Virtualiza-tion Platform (NVP) [53] offers this abstraction without requiring any support from the underlying networkinghardware. The solution is use overlay networking to pro-vide each tenant with the abstraction of a single switchconnecting all of its virtual machines. Yet, in contrast toprevious work on overlay networks, each overlay nodeis a actually an extension of the physical network—asoftware switch (like Open vSwitch [57, 63]) that encap-sulates traffic destined to virtual machines running onother servers. A logically centralized controller installsthe rules in these virtual switches to control how packetsare encapsulated, and updates these rules when virtualmachines move to new locations.
Network virtualization for evaluating and testing SDNs.The ability to decouple an SDN control application fromthe underlying data plane makes it possible to test andevaluate SDN control applications in a virtual environ-ment before the application is deployed on an opera-tional network. Mininet [41, 48] uses process-based vir-tualization to run multiple virtual OpenFlow switches,end hosts, and SDN controllers—each as a single pro-cess on the same physical (or virtual) machine. The useof process-based virtualization allows Mininet to emu-late a network with hundreds of hosts and switches ona single machine. In such an environment, a researcheror network operator can develop control logic and easilytest it on a full-scale emulation of the production dataplane; once the control plane has been evaluated, tested,and debugged, it can then be deployed on the real pro-duction network.
•Virtualizing (“slicing”) an SDN.In conventional net-works, virtualizing a router or switch is complicated,because each virtual component needs to run own in-stance of control-plane software. In contrast, virtualiz-ing a “dumb” SDN switch is much simpler. The FlowVi-sor [68] system enables a campus to support a testbed fornetworking research on top of the same physical equip-ment that carries the production traffic. The main ideais to divide traffic flow space into “slices” (a concept in-troduced in earlier work on PlanetLab [61]), where eachslice has a share of network resources and is managedby a different SDN controller. FlowVisor runs as a hy-pervisor, speaking OpenFlow to each of the SDN con-trollers and to the underlying switches. Recent workhas proposed slicing control of home networks, to al-low different third-party service providers (e.g., smartgrid operators) to deploy services on the network with-out having to install their own infrastructure [89]. Morerecent work proposes ways to present each “slice” ofa software-defined network with its own logical topol-ogy [1, 22] and address space [1].
网络虚拟化（根据逻辑网络对物理网络的抽象）显然不需要SDN。同样，SDN（逻辑集中的控制平面与基础数据平面的分离）并不意味着网络虚拟化。有趣的是，网络虚拟化和SDN之间已经出现了共生关系，这已经开始催生了几个新的研究领域。 SDN和网络虚拟化涉及三种主要方式：
SDN是一种用于网络虚拟化的支持技术。由于云提供商需要一种允许多个客户（或“租户”）共享同一网络的方法，因此云计算使网络虚拟化成为重要问题。 Nicira的网络虚拟化平台（NVP）[53]提供了这种抽象，而无需底层网络硬件的任何支持。解决方案是使用覆盖网络，通过连接所有虚拟机的单个交换机的抽象为每个租户提供数据。但是，与以前在覆盖网络上进行的工作相反，每个覆盖节点实际上都是物理网络的扩展-一种软件交换机（例如Open vSwitch [57，63]），用于封装发往其他服务器上运行的虚拟机的流量。逻辑上集中的控制器将规则安装在这些虚拟交换机中，以控制数据包的封装方式，并在虚拟机移至新位置时更新这些规则
用于评估和测试SDN的网络虚拟化。将SDN控制应用程序与底层数据平面分离的能力使在虚拟环境中测试和评估SDN控制应用程序成为可能，然后再将其部署在运营网络上。 Mininet [41，48]使用基于进程的虚拟化来运行多个虚拟OpenFlow交换机，终端主机和SDN控制器-它们作为同一物理（或虚拟）计算机上的单个进程运行。基于过程的虚拟化的使用使Mininet可以在一台计算机上模拟具有数百个主机和交换机的网络。在这种环境下，研究人员网络运营商可以开发控制逻辑，并在生产数据平面的全面仿真中轻松对其进行测试；一旦对控制平面进行了评估，测试和调试，就可以将其部署在真实的生产网络上。
•虚拟化（“切片”）SDN。在传统的网络中，虚拟化路由器或交换机非常复杂，因为每个虚拟组件都需要运行自己的控制平面软件。相反，虚拟化“哑” SDN交换机要简单得多。 FlowVi-sor [68]系统使校园能够在承载生产流量的同一物理设备之上，支持用于网络研究的测试平台。主要思想是将流量流空间划分为“切片”（在PlanetLab的早期工作中引入的概念[61]），其中每个切片都共享网络资源，并由不同的SDN控制器进行管理。 FlowVisor作为hyper-visor运行，向每个SDN控制器和底层交换机讲OpenFlow。最近的工作提出了对家庭网络的切片控制，以允许不同的第三方服务提供商（例如，智能电网运营商）在网络上部署服务，而无需安装自己的基础设施[89]。最近的工作提出了一种方法，以软件定义的网络的每个“切片”呈现其自己的逻辑拓扑[1，22]和地址空间[1]。
参考文献：The Road to SDN: An Intellectual History of Programmable Networks