The research for a new paradigm for a better architecture of networking began in 2004. After several proposals and researches, the concept of software defined networking (SDN) emerged in 2008. In 2011, Open Networking Foundation was formed to take this concept to an executable level. The foundation consists of 69 members including Cisco, Juniper, HP, Dell and IBM.
Google, Yahoo, DT, Microsoft and Facebook serve as the Board of this foundation. One of the first SDN projects was AT&T’s GeoPlex. Some of the available commodity switches that are compliant with OpenFlow protocol of SDN are Hewlett-Packard (8200zl, 6600, 6200zl), Brocade (5400zl, 3500/3500yl) and IBM NetIron CES 2000 Series.
In traditional networks, most of the network functionality is implemented in hardware devices like switches and routers. In such systems, each appliance needs to be configured individually. Moreover, a router will treat all incoming packets equally and will follow the same protocols for all.
Inculcating such functionality is under the control of the device provider. So when a new technology, feature, update or application is introduced, all hardware devices need to be replaced. Thus, it has been found that traditional networking evolves slowly, is not dynamic and has limited functionality, as provided by the vendors. Being hardware-centric, it proves to be a huge challenge when the network is big and complex.
Software defined networking can be treated as a completely opposite architecture. Devicem level management is eliminated from networking. Here, the network administrator has the authority to shape the traffic from a centralised control console without configuring individual devices in the network. Unlike traditional networks, where the switch forwards all frames as per a pre-defined program, software defined networking has the provision to send the decision to switch through an application running on a server. Also, the switch can query the controller to provide information about the current traffic. This results in a highly-scalable, flexible and agile network.
Need for new architecture
There are several reasons why the traditional hardware-centric architecture needs to be replaced.
Increasing number of users
As shown in Fig. 1, advancement in technologies is leading to a drastic increase in the number of Internet users. Technologies and devices are becoming user-friendly. These have now become an inevitable part of our daily lives. This calls for a demand of upgrade in the technology to accommodate the increasing traffic.
Continuously varying traffic pattern
The number of users at any particular place is not always the same. Traffic pattern, thus, keeps changing continuously. There is a drastic change in the number of users at day time than at night time. Due to such variations in traffic pattern, there arises a demand for a flexible technology that should be capable of handling such traffic.
To eradicate vendor dependence
Traditional networking devices are functioned and programmed by technicians. The network administrator cannot modify the program or update the protocol. Thus, when any new algorithm or protocol is to be implemented in the network, altogether, the existing devices become useless.
Evolution of cloud technology
Companies are now moving towards cloud computing. Most of their data is stored on cloud servers for easy and quick access. Ultimately, it increases the burden on network architecture and increases traffic.
Increasing number of networking devices connected through the Internet
As shown in Fig. 2, researchers predict from the statistics that the number of connected networking devices may reach 50 billion by 2020.
Fig. 3 reveals the complexity of the network. To manage and control such a tremendous volume of traffic, appropriate architecture is required.
Evolution of the Internet of Things (IoT). As shown in Fig. 4, the IoT may connect 50.1 billion networking devices by 2020. The IoT market revenue is expected to grow from US$ 1.928 billion in 2013 to US$ 7.065 billion in 2020. Such a rapid evolution results in the demand for highly-reliable and agile networking architecture. Data-intensive services offered by the IoT like video surveillance, emergency services and uninterrupted monitoring demand a technology that supports the cause with excellent quality of service and high data rates.
Architecture components of SDN
Fig. 5 shows the various components of SDN architecture. The architecture comprises three planes, namely, data, control and application. Data plane includes the dedicated hardware and is responsible for the processing of packets as well as traffic forwarding.
In the controller plane, policies are configured to define the scope of control given to SDN application and to monitor system performance. Optimising resource allocation and checking network failure are also the roles of controllers. In both planes, security functions are considered to ensure secured intercommunication.
As demonstrated in Fig. 6, in traditional networking, each device has an embedded set of protocols that make the routing decision. Whereas in SDN (Fig. 7), there is a central network operating system (OS) that controls all devices in the network. All decision-making protocols and access-control algorithms are centrally managed by the control plane.
OpenFlow protocol
OpenFlow is an application program interface (API) that provides interface between data and the control plane. It controls the forwarding table of the router from a remote location. OpenFlow based controllers maintain an inventory of all paths of the network and store these for further allocation. The controllers update the flow table as per network traffic, using OpenFlow protocol.
Northbound
In general terms, northbound interface allows network components to communicate with higher-level components of the same network. Here, northbound is the API that interfaces control plane with business application plane.
Southbound
Southbound interface allows network components to communicate with lower-level components of that network. In SDN, OpenFlow protocol behaves as southbound, enabling intercommunication of controller and data planes. Extensible Markup Language (XML) and Lisp are other examples of southbound SDN.
Implementation considerations
Three basic aspects that are taken into prime consideration while implementing the new architecture are:
Quality of Service (QoS)
Considering the current trend of growing Internet users, the prime factor for acceptance of any new technology is the QoS it provides. The traffic-handling capacity and error rates are the parameters that judge the efficiency of any technology. Protocols for dynamic networking are needed to be designed such that there is no compromise with the QoS it offers.
Security
SDN security requirements differ from those of a static network due to their inherent characteristics. Protocols are expected to reduce the security threats by deploying SDN controllers within their secure computing environments. Being dynamic, SDN is more prone to hazardous security attacks. However, the networking environment needs to be scalable, secured and efficient, simultaneously.
Flexibility
Adaptation of new technology is possible only if its implementation has flexibility. The objective behind switching to a new technology can be achieved by making the technology user-friendly.
Also, scalability is a mandatory characteristic of any network, especially when it is a dynamic architecture as more number of private networks join the existing network every day (the IoT).
Benefits of SDN
Better traffic-handling capacity
In today’s networks, proprietary firmware on the switch determines where frames are forwarded. In software defined networking, network administrators can actually shape network traffic. They can do this from a centralised network console that integrates the information and controls all their network switches into a kind of network fabric. They can also change data traffic rules on-the-fly if they need to.
The network administrator has complete control over network traffic through a software interface that software defined networking provides. This allows organisations to decrease their reliance on more expensive switches with proprietary firmware that perform these functions—and that must be set manually.
Relaxing vendors from the burden to embedding specific functionalities into individual devices
Since vendors do not need to embed any single function/program into individual devices, their task becomes easier and time saving. A common prototype of a device can be approved, and all devices must be manufactured such that these are flexible and programmable.
Central management
SDN provides network administrators an opportunity to make their network device adjustments through a software interface instead of manually configuring individual hardware. Fig. 8 shows the comparison between traditional networking and SDN. It is clearly displayed that, in SDN, no separate network OS is available with individual devices. Rather, a central OS controls each and every connected device, and all features are generalised and not specific for individual device.
To summarise, owing to the increase in the use of the Internet and evolution of new technologies like cloud computing and network virtualisation, there is need of a new architecture that can offer flexible and affordable services.
Software defined networking is a promising architecture that removes manual configuration and hardware dependence. It is totally software centric, and the authority to configure and implement different protocols and handle traffic is provided exclusively to the administrator. This promotes efficient use of bandwidth and better handling of traffic. Networks would no longer be a mere collection of various hardware devices.
Feel interested? Check out other electronics projects.
