The beauty of sdn lies in its ability as a technology to make networks flexible, ensure efficient use of resources, and facilitate a much higher level of system autonomy. Like any nascent technology, sdn should be handled cautiously to avoid it becoming

an attack vector. However, sdn opens up new possibilities for the implementation of improved security mechanisms in the network, offering broader visibility, programmability, as well as a centralized approach to network management.

So we can say that at this moment the last issue about SDN is security.

best regards.

Hi Abdullah, 

I would say SDN is VERY much still a research field. In fact, current commercial and let us say "open-source" development driving what is called "Software-defined networking" relies heavily on manually configuring orchestration procedures through a number of interlinked platforms to instantiate and run services and overlay technologies which run over existing lowering level networking technology.

However, SDN, at its truest predates what we currently see in the market. Where SDN in a self-aware sense runs unsupervised and dynamically learns about it's topology, resources and environment. With current SDN these behaviours must be pre-configured into the network, rather than the network making it's own "learned" decisions.

So, I would suggest here the "last" isn't the last, because in actuality SDN is only just revealing new areas of study for networking that in the academic space have been around for quite some time. That is, machine learning, or namely neural networking.

Here are a few papers from the famous Erol Gelenbe who arguably created the first "Real" SDN:

"Cognitive packet networks"

"QoS and routing in the cognitive packet network"

"Energy-aware routing in the cognitive packet network"

"The cognitive packet network: a survey"

"Design and performance of cognitive packet networks"

"Demonstrating cognitive packet network resilience to worm attacks"

"Recursive routing in the cognitive packet network"

This may not directly answer your question, but I hope it expands your thinking somewhat on less conventional SDN as it is being sold currently!

Thank you for answering. So, you mean that there is no comertial system work on SDN yet or isn't brought to life yet. That is what I ask about.

If we are referring to SDN in relation to a pure unsupervised and self-configuring network (e.g., artifical intelligence), then no, none outside the academic space have been made widely accessible for enterprise and production systems. Although, they may be in development.

There may be others that argue the definition of SDN. It is not the most definitive description of a network. You might argue that Self-aware (AI) is or isn't SDN. Commercial marketing terms differ from those in the academic space.

However, currently, what we have are a series of network overlay (think VXLAN) and virtualisation (think Cisco NEXUS 100v or Arista EOS) technologies that are systematically connected via SDN management servers that provide orchestration and provisioning of services, such as changing the network topology and functionality dynamically and on-demand based through pre-configured and let's say "bolt-on" machine learning algorithms that are being fed information about the network. In reality, the machine learning tools that are being used to generate self-configuring behaviour in commercial SDN systems are not existentially built in to the fabric of the network (e.g., inside a transport protocol). Unlike the example I gave "Cognitive Packet Network" using the "Random Neural Network" algorithm.

Maybe the following article gives u some ideas on what they're doing out there in academia (practical work). 

https://www.researchgate.net/publication/286921892_OFTEIN_A_Community_Efforts_towards_OpenShared_SDN-Cloud_Virtual_Playground

I think there might be investments in SDN by some companies but for sure not the SMEs, perhaps due to the high implementation cost and changing the whole structure of the network which works nicely for them.

Conference Paper OF@TEIN: A Community Efforts towards Open/Shared SDN-Cloud V...

I offer you this paper :

http://ieeexplore.ieee.org/xpl/login.jsp?tp=&arnumber=6994333&url=http%3A%2F%2Fieeexplore.ieee.org%2Fxpls%2Fabs_all.jsp%3Farnumber%3D6994333

and this:

http://dl.acm.org/citation.cfm?id=2486019

In second paper you can read about google's experience with SDN in deploying their data centers.

SDN appears to be a known concept which has been rejuvenated in the last 8-10 years. In a way, Intelligent neworks as defined in early '90s embedded many of the features that are now developed in the new Software-Defined Networking area. One of the reaons for SDN success is that these concepts have been now endorsed by the industry, not only in the telco world but also in the IT world as it becomes embedded to cloud computing technologies.As some previous experts wrote, SDN allows networks to be programmable and then more flexible. There are many initiatives and platforms implementing SDN features. Though SDN in wide area networks are still to come at scale, there are now various implementations of SDN in the data center environments. Additionally, convergence between IP and optical networking is also evolving towards Transport SDN. Meanwhile with the softwarization of networks mainly based on virtualization technique, it appears that SDN is a piece of a large field that encompasses Network Function Virtualization (NFV) and cloud architectures. As such, one can expect that the overall industry will move towards this softwarization that may impact all vertical applications of Digital Society.

So there is certainly a lot of issues of current interest in SDN (at large) that include control architecture, placement, distribution, resilience and security. Application fields are also developing quite fast so we can expect that interest in SDN will still grow in the next few years.

For additional info, please visit the IEEE SDN Initiative website at http://sdn.ieee.org/

I think that it depends on the framework you conceive SDN. In terms of network programmability, SDN provides an easy and unified way to perform network management. However, currently SDN still remains a hot topic as Prosper Chemoul previously mentioned, because it is considered as an enabling technology for NFV (Network Function Virtualization). So, according to this trend, which is currently one of the hottest research topics across various industries, the virtualization/softwarization of HW-based appliances is based on the framework of the Virtual Network Function (VNF), which encompasses the flexibility and the advantages of the softwarization and cloudification, where the SDN is exploited for performing the necessary network engineering towards efficiently acheving traffic steering via the deployed VNFs.

Practically speaking, the trend is the replacement of network HW-based systems with software ones hosted in clouds (significant CAPEX and OPEX reductions, while flexibility and adaptability is better), where SDN is used for forwarding/steering the service traffic within the clouds and specifically through these softwarised systems: the VNFs (similarly as it happened when they were hardware-based). I recommend to have a look at the ETSI NFV MANO initiative, which will provide you sufficient information on the SDN/NFV concepts and management/orchestration issues.

Furthermore, a closer look to ONF, will provide you also further aspects on SDN utilization after the deployment phase of the NFV, where SDN controllers are considered (usually in a  recursive manner) for managing the deployed Virtual Network Functions (VNFs).

There are also H2020 European Projects working in the field, such as T-NOVA, UNIFY, SESAME, VITAL, CHARISMA etc. They have produced many documents and results that will help you to understand what is trending now.