Software-defined networking

Software-defined networking
Name	Software-defined networking
Inventor	Martin Casado, Nick McKeown, Scott Shenker
Inception	2008
Related concepts	Network functions virtualization, OpenFlow, Cloud computing

Software-defined networking. It is a paradigm in computer networking that enables the network to be intelligently and centrally controlled, or 'programmed,' using software applications. This approach decouples the network control and forwarding functions, allowing network administrators to manage traffic from a centralized console without needing to touch individual switches. It represents a significant shift from traditional network architectures, promising greater automation, flexibility, and efficiency in managing complex data center and wide area network environments.

Overview

The foundational idea emerged from research at Stanford University and University of California, Berkeley, notably driven by Martin Casado, Nick McKeown, and Scott Shenker. It addresses the limitations of conventional networks where control logic is embedded in proprietary firmware of devices from vendors like Cisco Systems and Juniper Networks. By centralizing intelligence in software-based controllers, it abstracts the underlying infrastructure for applications and network services. This model facilitates innovative network management and has been widely adopted in cloud computing platforms such as Amazon Web Services, Microsoft Azure, and Google Cloud Platform.

Architecture

The architecture is typically structured into three distinct layers. The **infrastructure layer** consists of physical or virtual network switches and routers that handle packet forwarding. The **control layer** is the brain, featuring a centralized SDN controller like OpenDaylight, ONOS, or proprietary systems from VMware (NSX) and Nokia. This controller communicates with the infrastructure using southbound APIs. The **application layer** hosts business applications, such as those for load balancing or security, which communicate their requirements to the controller via northbound APIs. This separation is a core tenet, enabling programmability and dynamic configuration.

Key technologies

Several critical technologies enable its operation. The OpenFlow protocol, developed by the Open Networking Foundation, is a pioneering southbound interface standard that allows the controller to direct packet flow in network switches. Network functions virtualization complements it by virtualizing network appliances like firewalls and deep packet inspection tools. Representational state transfer APIs are fundamental for northbound communication, while overlay protocols like Virtual Extensible LAN create logical networks on top of physical underlays. Controllers such as those from the Linux Foundation's projects and commercial offerings from Hewlett Packard Enterprise are also pivotal.

Applications and use cases

Its programmability unlocks numerous applications across various domains. In data centers, it enables dynamic traffic engineering and efficient load balancing for giants like Facebook and Google. For service providers, it aids in bandwidth management and creating virtual private network services. It is crucial for implementing zero trust security models, allowing micro-segmentation and rapid policy enforcement. Other use cases include optimizing wide area networks via SD-WAN, managing Internet of Things device networks, and supporting the agility required for 5G core networks deployed by operators like AT&T and Verizon Communications.

Standards and protocols

A robust ecosystem of standards and protocols governs its implementation. The OpenFlow specification remains a cornerstone, maintained by the Open Networking Foundation. The Internet Engineering Task Force has developed related protocols such as ForCES and NETCONF. For network programmability, the IETF also standardizes YANG data modeling. Industry consortia like the Open Networking Foundation, European Telecommunications Standards Institute, and MEF Forum work on frameworks and interoperability. Protocols like BGP and MPLS are often integrated within architectures, while newer interfaces are defined in projects like OpenConfig.

Challenges and limitations

Despite its advantages, adoption faces several challenges. Security is a primary concern, as the centralized SDN controller presents a single point of failure and a lucrative target for attacks, potentially compromising entire networks like those of financial institutions. Interoperability between equipment from different vendors, such as Arista Networks and Cisco Systems, can be difficult. Performance and scalability of the control plane, especially in massive networks like those of national research and education networks, require careful design. There is also a significant skills gap, as network engineers must learn programming languages and new operational models, shifting from traditional CLI expertise.

Category:Computer networking Category:Network architecture Category:Cloud computing