SDN: concept changes over 5 years

Since 2011-2012, they began to talk a lot and write about the need to change the architecture of computer networks for a number of reasons. First of all, the cause was the avalanche-like growth of traffic and the change in its structure and the complexity of network scaling. It was during this period that video traffic began to grow exponentially: from 2010, Cisco predicted a 26-fold increase in video traffic by 2015. Mobile traffic grew at an incredible pace: from 2005 to 2015, 4,000 times. To the list of reasons, you can add the growing popularity of cloud architectures, the need to reduce the cost of ownership of networks, reduce dependence on vendors, etc. The most obvious successor to the traditional network architecture, where the Control Plane and Data Plane are combined, was called the Software Defined Networking (SDN) architecture, in which the network control plane was imposed on a dedicated controller. In the end, as the main advantage of SDN, economy and independence from the vendor were fixed.

The fact that SDN was at the center of a discussion about the future of computer networks, in our opinion, was influenced by three events:

- purchase of VMware by Nicira - the first SDN startup - for a record $ 1.26 billion for the industry
- the creation in 2011 of the Open Networking Foundation - a non-profit organization whose main task was to popularize a new approach to computer network management - SDN. The initiators were 6 companies - owners of the world's largest networks - Deutsche Telekom, Facebook, Google, Microsoft, Verizon, and Yahoo. Very quickly all the major manufacturers of network solutions joined the consortium, including Cisco, Brocade, NEC, HP, and so on.
- A report from Urs Hölzle , Google’s senior vice president of technology infrastructure, at the Open Networking Summit 2012 that the company uses SDN technology.
')
In the same 2012, promising marketing research began to come out in batches that the market for SDN solutions would reach enormous size in five years, and almost all of them called 2016 the year of “mass adoption”.

Actually, the very year 2016, which was supposed to be the “year of introduction”, was almost over and there is an opportunity to assess how much the predictions came true and what changed in general during this time.

1. If in 2011-2012 about SDN, many vendors and integrators (especially in Russia) were treated with great skepticism and as a university toy, by 2016 their opinion had changed dramatically. In 2011-2012, they stated that all modern equipment so everything is able and works perfectly, now all the players in the industry have come to terms with SDN and perceive this approach as our present. The same service providers like the fact that they can manage their equipment from the logical center, they like the fact that they can create the equipment management services and programs and recognize that such an approach is the future.
   
   
2. If in 2012, SDN was positioned primarily as a technology for data centers, and often even put an equal sign between SDN and OpenFlow (as a result, by the way, many articles appeared on what to compare, let alone put an equal sign between them, for example, this or this ), now the scope of SDN has expanded significantly. First, they added carrier networks. Mainly, SDN in telecom began to seriously think in 2015, when the AT & T telecom operator changed the paradigm and in fact declared itself a software company. By 2020, AT & T set a goal to virtualize and control more than 75% of the network using SDN technologies. And both software and hardware company plans to develop independently. Already, about SDN and NFV in the networks of service providers speak and write much more than they once wrote about SDN and data center.
   
   In addition, SDN solutions and cases for corporate networks have emerged - Enterprise SDN, where SDN is used in private clouds, in data processing networks, for Internet security, and more. In corporate networks, new technologies are being introduced more slowly, but some applications, such as, for example, SD-WAN (Software-Defined WAN), are already becoming common in large enterprises, and the number of variants of such solutions is constantly growing. SDN technology continues to evolve, engineers are not so skeptical about it, and it can be said with confidence that the number of implementations will increase in the next few years.
   
   
3. Over the past 5 years, a huge number of mergers / acquisitions and promising investments have occurred on the market. The very promised vendor independence gave hope for the emergence of start-ups in one of the most difficult areas in IT - network technologies. Even 15-20 years ago, the dictates of IT giants Cisco, Brocade, Juniper, etc. seemed unshakable. SDN promised a significant redistribution of the market. Our colleagues from NFware calculated that only in 2015, the SDN and NFV sphere attracted over $ 600 million of investment .
   
   
4. One way or another, all big vendors have got their own lines (or platforms) that will allow to program the network to one degree or another.

What did not come true:

• for sure 2016 will not be a year of mass implementation. Projections for the development remained all the same promising, but now they say about 2020-2022;
  
  
• many promised the death of OpenFlow. In fact, the protocol has not died and is still most ready for use in real networks and there is even a list of vendors from which you can buy equipment with support for OpenFlow version 1.3.4. But the advantage to it appeared a few more like the Protocol Oblivious Forwarding (POF) from Huawei or the P4 (Programming Protocol-independent Packet Processors) from the Open Networking Research Center (Stanford). In fact, OpenFlow is still the only SDN implementation that promises fair independence from the manufacturer. Yes, all large vendors offer their SDN solutions and to some extent expanded the ability to control the device on the network, but the solutions of different vendors are also practically incompatible with each other.

What we have now:

• no one argues that SDN is necessary, and all disputes have shifted towards realization;
  
  
• in fact, now everyone has understood what to expect from SDN both in terms of implementing SDN / OpenFlow in practice, and in terms of the emergence of new equipment, and in terms of the emergence of new services. Exactly the SDN application areas will not change. For example, it can be argued that as the OpenFlow switches did not appear during this time, they most likely will not appear, since as long as large vendors do not change the production of their equipment, there will be no progress in this area. Most likely the OpenFlow protocol will slowly evolve in the direction that suits both vendors and their clients.
  
  
• all vendors have a completely different understanding of SDN. Completely different protocols, different approaches to implementation, one way or another centralized management appeared, it is possible to independently create services, etc., but there is no openness and portability between devices. But since 2015, all major SDN vendors have begun to actively build horizontal links to create complex products that end customers want.

In fact, instead of the stage of mass implementation, we are now witnessing the stage of mass piloting the technology. All large telecom operators are targeting technology, selecting teams that will be able to work with new technologies, planning restructuring of business processes. We think that in the next 2-3 years it will be possible to talk about the beginning of the implementation stage.