This vendor-written tech primer has been edited by Network World to eliminate product promotion, but readers should note it will likely favor the submitter’s approach.
As enterprises add mobile users, virtualized services, and public and private clouds, they're running into constraints that threaten to limit their ability to scale their network infrastructure appropriately. Contrary to what many assume, the constraints are not so much bandwidth and cost, the real culprit is complexity.
Software Defined Networking (SDN) and Network Function Virtualization (NFV) address the complexity of 21st century computing with an agile, centrally managed, automated network that can grow, scale, and change much more easily and effectively than today’s static network architectures. Deploying SDN and NFV in the branch office allows IT to get familiar with the technology before scaling it across the rest of the network.
Drivers of Network Complexity
Network complexity has increased dramatically as a result of several trends:
* Mobile devices and applications - Administrators have been hard pressed to achieve the visibility, management and security necessary to accommodate thousands of laptops, smartphones, tablets and other new mobile end points without endangering sensitive information and network performance.
* Virtualization - Virtualization (particularly server virtualization) has caused the number of servers in the average organization to skyrocket. The reason: Provisioning new virtual machines takes minutes compared to the weeks or months it takes to purchase server hardware. Since they’re so easily deployed, virtual servers are often added without the knowledge of IT, making the task of monitoring, managing, troubleshooting and scaling IT infrastructure that much more difficult. Virtualization also adds its own layer of management complexity.
+ ALSO ON NETWORK WORLD: SDN FAQ +
* Cloud Computing - With virtualization has come automation, orchestration, user self service and the public and private cloud-- turning what was once a static, fenced-off IT infrastructure into an ever changing, ever scaling, growing, shrinking, morphing, distributed beast. The advantage has been unprecedented agility and fast time to market for new products and services. The drawback has been complexity. Gaining visibility into such a dynamic infrastructure has led to a proliferation of system, network, and other logs used to troubleshoot issues that inevitably come up and catch small problems before they become big ones.
The Problem with the Static Network
Unfortunately, unlike the automated, virtual world of servers and the cloud, today’s networks still run primarily on thousands of specialized, proprietary routers and switches that require a lot of manual configuration using an archaic command line interface (CLI). While new services can be deployed on virtual servers and storage or in the cloud in minutes, they often have to wait a lot longer while necessary adjustments are made switch-by-switch and router-by-router to the static, distributed network.
SDN and NFV
The good news is that over the next several years large portions of the static, distributed legacy network will likely be replaced by Software Defined Networking (SDN), an evolving network architecture that decouples network control from thousands of specialized, distributed network switches and routers and moves it into centralized software. If SDN plays out the way it’s intended, future networks will be configured centrally via intuitive graphical software rather than switch-by-switch and router-by-router using the arcane command line interface (CLI).
Centralized control will make configuring new services and their network requirements and adjusting to network congestion and other challenges a much easier and faster process throughout the network. Centralized management also promises to help IT achieve better visibility into the entire network than it has had until now, so IT will be able to monitor and adjust networks more successfully as they scale and change to accommodate new circumstances and services. The CLI will still be available, but used only when engineers need a highly granular diagnostics interface to dig deep into the bowels of complex networking systems in order to address more challenging network issues.
Automation will be the second benefit of decoupling network control from lower level network nodes, thanks to SDN’s northbound and southbound interfaces. Southbound interfaces allow SDN network control software to communicate with lower level network nodes, including network and virtual switches and routers. OpenFlow is a standardized southbound interface under development whose purpose is to allow control planes to communicate with networks of multivendor hardware to discover topology, define network flows and implement requests coming in from the northbound interface.
The northbound interface is the interface between applications and the control plane. Its beauty is that it allows applications to tap into the control plane to make requests for bandwidth, quality of service, and other functions automatically. Not only will management applications harness the northbound interface for centralized network visibility and IT control, but applications such as VoIP will be able to draw the bandwidth, quality of service and other network services they need to fulfill their performance and latency requirements automatically. Ultimately, applications may provide network self-healing and adaption capabilities that automate network fault detection, isolation and remediation workflows under centralized policy control.
NFV works hand-in-hand with SDN, turning what was once dedicated network hardware into virtual machines that can sit on any server along with other applications and services. There was a time when server-based switches and routers couldn’t compete in performance and functionality with their specialized, dedicated, physical counterparts, but with advancements in the technology and more powerful servers used in the virtual environment, they’re catching up fast. NFV is becoming perfectly usable in all but the most demanding network environments.
The prevailing sentiment is to first deploy SDN and NFV in the data center. However, the combination of SDN and NFV also allows initial deployment in the branch office where the routing service can just be a software component on general purpose servers.
Allowing enterprises and MSPs to deploy to new branch offices rapidly and harness existing SDN management software to gain network visibility has enormous benefits. It allows the network engineers to get familiar with the technology in a less risky environment than deploying it directly in the core of a highly complex data center. This is much the same way enterprises started their initial VoIP deployments in the branch office before bringing it into the core. It also allows resource constrained branches to continue to use the same servers to be multi-purpose machines and avoid introducing costly dedicated switches to be separately managed and maintained.
As SDN and NFV proliferate, the static, complex configuring and reconfiguring of the network should become more virtual and automated, allowing network administrators to focus more of their time and resources on network architecture issues and service deployment rather than the painstaking configuration of distributed static boxes via an arcane CLI. This will allow networks to scale more rapidly and cost effectively, addressing the barriers the static network has created to growth. SDN management solutions will also provide more visibility for network configuration, monitoring and control. It will take time of course, as vendor products and standards mature and develop. But judging from the carriers and the enterprises, SDN and NFVs time has come.
David Corley is director of product management at Netsocket, a provider of virtualized, software-defined networking.