The SD-WAN track has changed. When will the dragon trainer appear?

In an environment where cloud computing, mobile applications, and enterprise globalization have become the mainstream, more and more real-time applications such as remote offices, video conferencing, and remote desktops need to be transmitted between multiple nodes. The contradiction between problems such as disconnection and slow access and the requirements of application immediacy and real-time performance has gradually emerged. At this time, SD-WAN came into being.

[[261951]]

What is SD-WAN?

SD-WAN, or software-defined wide area network, is a service formed by applying SDN technology to wide area network scenarios.

This service is used to connect enterprise networks, data centers, Internet applications and cloud services over a wide geographical range. Its typical feature is to 'cloudify' network control capabilities through software and support the opening of application-aware network capabilities.

Let's take a familiar example. A patient went to see a doctor because of internal heat. He didn't know what internal heat meant, but he knew the symptoms of internal heat. The doctor didn't know the specific cause of the internal heat, but he could tell what he was talking about. So the doctor asked the patient, where is the internal heat? The patient answered that he was internal heat everywhere. In the end, the doctor gave a solution: drink more hot water.

To put it simply, SD-WAN is a service disguised as network technology.

The formation of SD-WAN

Times make heroes. The Internet undergoes a major technological change every ten years or so. There are many driving forces behind the emergence of SDN and SD-WAN.

From the centralized control architecture and Openflow protocol proposed by Stanford University in 2006 to the commercialization of Google B4 in 2012, the focus of SDN in its embryonic and disillusioned stages has always been on cloud and data center scenarios, and the industry is eager to find the next killer app. On the other hand, the development of cloud computing technology has triggered changes in the IT industry, and Internet+ has driven the transformation of traditional industries. In 2014, the Internetization of B-to-C surpassed the attention of B-to-C.

The relationship between applications and networks has undergone subtle changes. The previous balance has been broken and a new balance is needed. Opportunities have emerged during the brief period of turbulence. It can be said that SD-WAN services focusing on the enterprise market and wide area network are formed in such an industry background and expectation.

Technical conditions such as the rapid growth of resource supply and the explosion of business diversity, as well as non-technical conditions such as the saturation of traditional market areas and strong demand in emerging markets, have led to the industrialization of SD-WAN services. Facing the personalized requirements of customers and mobilizing a variety of resources and services at the same time, SD-WAN is in the middle position of services, and only when it has both the above and below conditions can it generate value.

If we divide SD-WAN into user and application-oriented, government and enterprise-oriented, ICP-oriented, and ISP-oriented based on who it serves and where the resource scheduling comes from, this division is dynamic and will continue to evolve with the development of the entire network.

The three phases of SD-WAN

The development of SD-WAN in recent years can be divided into three stages.

The first stage: hype concept

The characteristics of this stage are: concepts are flying all over the place, no one can convince anyone else, and no one can understand what others are saying, just like the Tower of Babel in the Bible story. The result is to let the bullets fly for a while.

By 2016, everyone realized that resources are the most valuable in SD-WAN, and owning scarce resources is like having a mine at home. In 2019, operators began to use huge resources to impact the original track, and SDWAN began to transform into a service.

The second stage: the 2.0 stage where resources are king

At this stage, the main contradiction between the application layer and the network is insufficient bandwidth. Faced with this situation, TCP acceleration technology made its first fortune, but soon began to face the problems brought by the digital divide. At this time, companies of all sizes around the world are working to fill the digital divide.

Major Internet companies are increasing their investment in global data centers and transmission resources. Among them, for domestic companies, the driving force comes from the "Belt and Road Initiative". Against this background, China Mobile and China Telecom began to build their own government and enterprise networks. China Telecom built a large-scale optical cable network covering the Indochina Peninsula and entered the Philippines.

With the entry of SD-WAN2.0 Terminator, the track structure has changed. The era of relying on one line to dominate the market is over.

The third stage: 3.0 global service era

Entering the 3.0 era, from a domestic perspective, the market is saturated, but if we look at the world, there are only a handful of Chinese multinational companies in the true sense, which means that Chinese companies are facing a completely blank state in the entire overseas market.

At this time, we need to be concerned about two issues. One is, when the focus of the entire economic development begins to tilt overseas, should SDWAN be service-centric or network-centric? The other is the factors that jointly promote the emergence of the industry.

How to Become a Dragon Trainer

Since the 2.0 era, the SD-WAN track has changed. Under the "Belt and Road" policy, operators have become more powerful, and Internet giants have entered the market one after another, which can be regarded as the entry of dinosaurs. Startups that initially dominated the SD-WAN world are now competing on the same stage with large companies. How to choose the track has become very important.

To become a dragon trainer, you need two abilities: the art of OTT and network mapping.

Google is a good example of the art of OTT. In the past few years, all the new technologies it released covered the transport layer and the network layer, but all the technologies had one thing in common, which was that they were all based on not affecting the interests of operators. Whether it was B4, ESPRESSO, BBR, or QUIC, they all tried not to stimulate the underlying operators' networks as much as possible and maintained the expectation of continuous and stable output of the underlying network. This was the superb art of OTT.

The necessity of global network mapping is that mapping is the basis of all intelligent control. Taking autonomous driving as an example, the core part of autonomous driving is actually the annotation of maps and the collection of geographic information. Without these things, it is difficult to achieve autonomous driving by relying solely on the sensors of autonomous driving vehicles. These tasks are actually the basic work hidden behind various intelligence.

In fact, overseas resources are already very abundant. The problem now is no longer about renting a dedicated line to transmit something, but knowing where the required things are when pooling resources. When global network and computing resources are fully pooled, we see that almost all SDWAN architectures discuss the control plane and data plane, but not the management plane.

From a common sense perspective, a controller or control application actually relies on a mathematical model when making decisions. Whether this model can make good decisions does not depend on whether the model is complex and smart, but on whether the input data is accurate enough. Therefore, the management plane is the core of the SDWAN system, and the cost of the management plane is the cost center of the entire system.

In addition, we need to clearly know what users really need in overseas environments. In the future, more and more capabilities that were previously closed in black boxes will be released and turned into cloud-based resources. At this time, relying solely on static resource information cannot provide sufficient decision-making basis, and we must rely on dynamic resource information. Rapid deployment, flexible deployment, and on-demand deployment become the key. Startups are not Swiss Army knives. High specialization and providing services in niche areas are in line with the development laws of the entire industry.

When all resources of the entire Internet can be clouded, pooled, and integrated, a new era will come. Before that, it will take time to accumulate a lot of work and wait for the accumulation to be successful.

The second half of SD-WAN: Walking with Dinosaurs, When Will the Dragon Trainer Appear?