Managing the Abstract Network in the Era of SDN
How far are we from self-service networks of the future?
Software defined networks (SDN) and advanced fabric topologies represent the state-of-the-art in data communications, but all of this scalability and flexibility will be of little use if the emerging network architectures become too complex.
Already, the app-centric approach to infrastructure management is putting more control in the hands of non-technical users, which means any abstract networking environment will require intuitive interfaces and a higher degree of automation than what we have now. But how are these constructs supposed to deal with issues like resource contention and systems availability if everyone who wants to deploy an app is leveraging the environment to their own ends?
What’s needed, says technology journalist John Edwards, is a new approach to networking that is more in tune with the service-driven workflows that characterize the digital economy. Today’s network architectures were designed in the x86 era of client-server architectures and are completely out of tune with the dynamic provisioning and other requirements of today’s enterprise. With SDN as a base, organizations will need to do away with the closed networking and device-level management of the past in favor of more open system that stresses programmability and pooled resource allocation.
While this type of functionality can certainly be layered onto legacy network infrastructure, new hardware platforms are starting to focus on the need for greater abstraction as well. Brocade’s new SLX 9850 router, for instance, features the company’s Insight Architecture that marries real-time visibility with the Workflow Composer system to support server-based, DevOps-style automation. In this way, the enterprise can support high-speed, highly scalable deployments while maintaining operational support of increasingly diverse workloads as they traverse multiple networking domains in the data center and the cloud.
But these types of solutions still leave much of the network management responsibility in the hands of trained professionals. How much better would it be if the basic provisioning and allocation tasks could be performed by users, allowing network techs to concentrate on higher order architectural and strategic development projects? Researchers at the University of Texas at Arlington are working on just such a platform, fueled by a generous gift from the National Science Foundation. The objective is to develop a model that identifies and implements precise resource configurations in response to user-defined service-level requirements. In this way, application stakeholders can simply express what they want to accomplish and the IT management stack takes care of the rest – invariably in ways that are more effective and less costly than traditional manual provisioning.
The reality of today’s data environment, however, is that very little of it is maintained on the enterprise’s private network. With cloud computing evolving into device-driven IoT infrastructure, coordination of network resources will have to extend across multiple third-party networks. Companies like wireline carrier Windstream are looking to bring automation to managed wavelength services, allowing organizations to tailor their bandwidth consumption more closely to application needs. The company has deployed Ciena’s Blue Planet orchestration platform to convert its optical network into a set of programmable resources that can be parsed into 1, 10 and 100 GbE services. With its open architecture and support for third-party SDN controllers, Blue Planet increases network speed and flexibility across long-haul and metro networks, and it lays the groundwork for full traffic orchestration across multiple technology domains, both on physical and virtual architectures.
Network management will never be easy, but with enough abstraction and automation it can be done in ways that mask the complexities while allowing knowledge workers to focus on results. But this level of functionality does not happen by itself simply because networks are defined in software. Real network productivity only comes about when the management stack is sophisticated enough to deliver the resources that people need at costs the enterprise can afford.
Arthur Cole covers networking and the data center for Enterprise Networking Planet and IT Business Edge. He has served as editor of numerous publications covering everything from audio/video production and distribution, multimedia and the Internet to video gaming.