SDN (Software-Defined Networking), NFV (Network Function Virtualization), and NV (Network Virtualization) are hot topics in both the telecommunications and IT industries. These concepts are closely tied to carrier networks, cloud computing, and virtualization, and they often appear in different contexts, leading to confusion among many people. In this article, I will share my understanding of SDN, NFV, and NV based on my previous studies and experiences.
When discussing SDN, it's important to note that it refers to a software-defined network. The Open Network Foundation (ONF) defines SDN by three key characteristics:
- Separation of the control plane from the forwarding plane
- Centralization of the control plane
- Programmability of the network
Those familiar with traditional networks can easily see that our current infrastructure is not SDN-based — it's hardware-defined. Whether it's carrier backbones, metropolitan area networks, access networks, data centers, or enterprise campuses, these networks have certain characteristics:
The control and forwarding planes are integrated within each device, making them specific, closed, and tightly coupled. The control plane operates in a distributed manner, where each device independently learns the network topology using Layer 2 and Layer 3 protocols to build forwarding and routing tables. Packets are then forwarded autonomously without centralized coordination.
Additionally, these networks are not programmable. Deployment typically relies on proprietary command-line interfaces from different vendors, which lack standardization.
In an ideal SDN environment, the forwarding plane would be generalized, allowing for more open and flexible deployment. Manufacturers could use general-purpose chips instead of specialized ones, reducing barriers to entry. The controller would manage the forwarding devices, sending instructions for packet handling.
The controller itself should be high-performance, capable of managing large-scale networks, deploying complex services, and ensuring high availability. It can even form clusters to support larger infrastructures. An open northbound interface would allow applications or cloud platforms to interact with the controller, enabling dynamic network service provisioning.
While there are various interpretations of SDN, the third feature — programmability — is often emphasized. However, I believe that without centralized control (the first two features), the benefits of programmability are limited. For example, without a global view of the network, configuration consistency and real-time status information cannot be guaranteed.
This is why the three core features of SDN are interdependent. Only when they work together can we truly break free from the limitations of hardware-defined networks, allowing higher-layer services to fully utilize the network as needed.
Google’s B4 and Facebook’s latest data centers are good examples of this approach. Their design emphasizes full control of the network through the controller, proving the importance of the first two features.
When talking about SDN, it's also important to mention OpenFlow, even though it may seem like just a southbound protocol. OpenFlow was introduced alongside SDN and plays a central role in its implementation. However, progress has been slow due to several challenges. First, current commercial chips struggle to support multi-field matching defined by OpenFlow. Second, the support for multi-stage pipeline processing is still limited. Lastly, interoperability between vendor-specific controller-switch solutions and existing networks is often too simplistic to meet real-world requirements, resulting in isolated deployments suitable only for testing rather than production environments.
Despite these challenges, OpenFlow’s efforts to generalize the forwarding plane are significant. If its standards are finalized, the development of OpenFlow-compatible chips could become as transformative for networking as x86 was for servers.
Now, let’s turn our attention to NFV. NFV, promoted by ETSI, aims to reduce the cost of network construction by replacing dedicated telecom equipment with general-purpose hardware and virtualization technologies. The initiative is backed by major players in the telecommunications industry.
Many in the IT community may not fully understand telecom networks and might think that NFV’s goal is to replace data center switches and routers with general-purpose servers. However, NFV covers much more than that. It includes fixed networks, mobile networks, and transmission networks, involving over a hundred types of network elements. This breadth makes it difficult for even experts to accurately count all the components involved.
Since NFV aims to replace dedicated telecom equipment with general-purpose servers, it's essential to understand their differences. Dedicated telecom equipment is known for high performance, reliability, and scalability. On the other hand, general-purpose servers offer strong computing power, standardization, and mature virtualization support.
Water-cooled Capacitor is supercapacitor is a capacitor with a capacity of thousands of farads.According to the principle of capacitor, capacitance depends on the distance between the electrode and electrode surface area, in order to get such a large capacitance, as far as possible to narrow the distance between the super capacitor electrode, electrode surface area increased, therefore, through the theory of electric double layer and porous activated carbon electrode.
Water-Cooled Capacitor,Water-Cooled Power Capacitor,Water-Cooled Electric Heat Capacitor,Water-Cooled Electric Heating Capacitor
YANGZHOU POSITIONING TECH CO., LTD. , https://www.yzpst.com