Software-defined networking (SDN), originally implemented in data centres, has become a popular discussion topic across the ICT industry. It has helped companies adjust network structures and configure virtual machines (VMs) to address constantly changing data centre requirements.
Pre-SDN, this had never been an easy task, especially for networks required to support multiple technologies such as mobile services, cloud computing, social media, big data, and the Internet of Everything (IOE). The absence of SDN made the delivery of real-time service transmissions, scalable mobile networks, and enhanced user experiences considerably more complex.
So, what is it, exactly, that SDN does? First, it decouples a network's control and forwarding functions such that the physical and logical networks can be treated independently. This decoupling allows network traffic flows to be programmatically reallocated efficiently to meet changing demands. SDN provides network administrators with a clear overview of the entire network via its centrally managed controller.
While often mentioned as a 'data centre' technology, SDN in fact has many different definitions and standards. Listed below are many examples of key opportunities and challenges that SDN is prepared to help address.
Opportunities
Enhancing efficiency of campus networks
While traditionally used in data centres to address frequent VM migrations, SDN can also be applied to campus networks. ICT vendors have conducted research and developed campus controllers for SDN-based campus networks, as wireless network users and mobile terminals are known to move around frequently and in ways similar to the movement of VMs in data centres.
Increasing bandwidth utilisation in wide area networks (WANs)
Traditional enterprise WANs operate at a 40% average utilisation over dedicated lines. When, for example, Google applied SDN technology to their internal WAN, their bandwidth utilisation jumped to more than 90%.
Huawei has worked with a data centre services provider to deploy SDN over its internal network. Having previously invested billions of dollars each year to achieve a bandwidth usage rate of only 30%, by upgrading to a dedicated WAN controller, its customers' WAN link usage rate increased to more than 90%, which in turn significantly reduced the rental cost of WAN links to their customers.
Customisation for specific industries
SDN protocols are based on open standards, and therefore, can be easily tailored to meet the requirements of individual companies or industry-specific technology vendors. Open standards allow the control policies of SDN-enabled hardware tailored to set specific domain-, rights-, and quality of service (QOS)-based policies. SDN networks can be customised on the forwarding-, device-, and network management system (NMS)-levels, as well as for controller-based application programming interfaces (APIs). Within the context of developing SDN solutions, secondary innovations can be implemented for defining policies for specific industry requirements in sectors such as healthcare, education and energy, among others.
Challenges
Support for smooth transitions to SDN
Many companies are currently looking for ways to adopt SDN, short of migrating their entire networks. These companies want to offer new, SDN-enabled services, but they also need guarantees that services running on existing networks will not be disrupted.
One solution for addressing this challenge is for companies to maintain their single physical network and add two logical networks: The first logical network will support existing protocols and services, while the second, the SDN zone, will be provisioned to support wholly new services. This logical split is achieved by using an Ethernet Network Processor (ENP) switch with dual control planes. ENP switches can be applied to support both network types in ways that ensure the smooth transition from traditional networks to SDN.
Adapting hardware for SDN deployment
OpenFlow is an industry-standard SDN communications protocol that allows servers to instruct network switches where to forward packets, and defines forwarding control interfaces, forwarding models and forwarding actions based on network services, and establishes flow tables for managing traffic flow.
Typically, traffic flow is managed with tables designed to support network device forwarding. For example, the Forward Information Base (FIB) table was developed for route forwarding, and the Media Access Control (MAC) address table for Layer 2 forwarding. These independent tables have been developed based on Application-Specific Integrated Circuit (ASIC) architecture and have predefined search orders for specific tasks. Average FIB tables support hundreds of thousands of traffic flows, while MAC address tables support a maximum of 20 000 traffic flows. However, these tables cannot support the application of OpenFlow in service networks, as OpenFlow requires larger tables for managing much larger traffic flows.
Network administrators can address the challenge of transitioning between network types by installing switches that include ENP chips. ENP chips are designed to support flow tables for both OpenFlow and traditional FIB and MAC services.
SDN has great potential and is one of the most promising emerging technologies in the networking space. It provides an open standards-based, agile, centrally managed, and programmable network for supporting the efficient and seamless service operations that are fundamental for companies competing in the highly dynamic business environments of the digital era.
Share