Where is the way out for SDN?

SDN was born in 2006. It was a campus innovation project led by Professor Nick McKeown of Stanford University. It proposed a programmable network architecture based on OpenFlow. After several years of technical research, it has gradually become commercially available in the past two years. SDN aims to improve network flexibility and enhance network deployment and operation efficiency. It broke the decades of silence in the network field, set off a whirlwind of network transformation, and quickly became commercially available.

[[250538]]

After years of market cultivation, we can see that there is sufficient technical awareness now. Every newly built data center will discuss whether to use SDN at the beginning. SDN has entered a period of rapid development. Of course, like other things, SDN also has some troubles of rapid growth, which even affects the development of SDN itself, making many data centers hesitant in choosing SDN. What is the way out for SDN?

SDN has experienced a period of rapid commercial development in the past few years, but it is now gradually slowing down. Many data centers are choosing whether to use SDN or not, and are also choosing the degree of virtualization. The benefits of SDN are self-evident, but the network is an important part of the data center, and it is also the most closed part. Sometimes, if the pace of transformation is too fast, there will be discomfort. SDN focuses on openness, and the design is a universal interface, which has allowed many people to discover opportunities, and many SDN startups have sprung up. However, it is not easy for these SDN software to truly manage network physical devices.

After decades of development, the network market has formed a market competition pattern dominated by Cisco, Huawei, Juniper, and HP. If SDN wants to truly control these devices, it needs to build a good relationship with these equipment vendors and adapt them well. Obviously, this is difficult to do. These mainstream equipment vendors are naturally unwilling to give up this piece of fat, not to mention that their size and strength are much stronger than those SDN innovation companies. These equipment vendors are also vigorously developing SDN, so how can they give those SDN innovation companies a chance? Therefore, the software of these SDN innovation companies is more of a tool for network monitoring and management. They manage the network by obtaining various operating information of the equipment through some open interfaces provided by the equipment. If they want to truly control the network, they are still a little short of being a network controller. Even the SDN software of these equipment vendors can only ensure that it is adapted to their own equipment. SDN has become a weapon for equipment vendors to shield their opponents when they conduct market competition activities. This has made the SDN market further closed, the market size has gradually become niche, the SDN market cannot be expanded, and in the end, it is still a few equipment giants who are playing SDN.

At present, SDN is still difficult to adapt across equipment manufacturers, and many data centers are unwilling to take this risk. Basically, if they decide to use SDN, the SDN software and hardware equipment must be from the same company. If they choose different companies, they will face adaptation problems. Data centers need to coordinate technology between multiple network suppliers, which is very labor-intensive. If they do not do this, there will be many unknown problems. However, if they choose all from the same company, the risk is not small. Once this equipment supplier has problems, it will have a fatal impact on the data center. Generally, more than two network suppliers are introduced, and the risk is controllable. This can also enhance their bargaining power, allowing several network suppliers to fully compete, and the data center will reap the benefits. The usual practice is that the digital center will select a small part of the network for SDN construction, and the other networks will still use traditional networking, so that they can see the performance of SDN and then deploy the SDN transformation pace of the future network. This is not a good choice for small data centers. The network is too small to be subdivided. Either the traditional network or the SDN network, one of the two.

The deployment cost of SDN is not high. You only need to introduce a few servers, install SDN software, and the network equipment supports Openflow. This also makes the real SDN market amount not large. Imagine that a few servers plus a software can only cost hundreds of thousands of yuan, which is very small compared to the tens of millions of network equipment. Even a large network can deploy a few servers with good performance. Therefore, SDN alone cannot make a lot of money. Moreover, in the early stage of SDN commercial expansion, software is mostly given away, and it can only be charged if it is used well, which also reduces SDN revenue a lot.

In this way, only those network equipment vendors who can provide overall SDN hardware and software solutions can make money, rather than just selling SDN software, which also makes life difficult for those SDN startups. It is cool to think that the entire data center network can be controlled with just a few servers and software, but if the control is too centralized, the safety of the entire data center network will be in the hands of the SDN controller, and the risk is not small. Once the SDN controller is attacked, the entire network will be paralyzed. This has attracted many doubts. Whether SDN can do a good job of escaping is related to whether SDN can be commercialized on a large scale. If the handling of SDN faults is not solved well, it is better to use traditional networks. Although it takes more manpower and material resources, it is safe to use. This is a mature model that has been developed over decades of networks.

On the other hand, if network control is not centralized on SDN, SDN only performs some configuration distribution and management, and monitors the network operation status, the role of controlling the network will be marginalized, and such network transformation is in name only and cannot be called an SDN network, but only improves the network management function. Therefore, it is difficult for data center customers to deal with SDN, which has also led to a slowdown in the deployment of SDN. Some data centers have delayed the introduction of SDN and entered a wait-and-see period to see how those data centers that have taken the lead in SDN perform before considering when to introduce SDN, which has greatly reduced the market opportunities for SDN.

It cannot be denied that SDN is still the direction of future network technology evolution. SDN makes the network open and intelligent. The difficulties are only temporary and are the growing pains that SDN technology must face in its development process. The development of SDN has reached a consensus in the technical and business circles. It is necessary to break the closed part of the network, and network equipment vendors have to invest in SDN.

Although SDN has encountered some difficulties for the time being, compared with the widespread commercial use of SDN two years ago, it has now entered a stable development period. Although there have been some doubts, they have not affected the way out. Most people are still optimistic about SDN. This is an excellent opportunity for SDN to change traditional networks. SDN still needs to move forward in great strides. There are more than one million data centers in the world. If all of them introduce SDN for network transformation, it will be a massive market with a market size of at least 100 billion. It is estimated that by the end of 2022, the SDN market size will reach 130 billion US dollars. SDN still has a way out. As long as it continues to develop and improve according to the established technology, SDN will still be wonderful.