Server and storage virtualization have given us a taste of what “dynamic workload management” can do for an enterprise. Now the appetite for this services-based business and operational model is growing – and rapidly. It’s time for the network to step up and feed the beast (so to speak).
Historically, most of the intelligence about the network has been housed in the network itself, shared between devices and extracted by management and monitoring systems for the purpose of understanding the current “state” of the network. But not much has been done with the information in terms of making the network smarter. With the advent of dynamic workload management, personalization and “bring your own device” (BYOD), the need to leverage, and in reality, wrest the embedded intelligence from the underlying silicon is becoming much more important.
As we build more highly distributed, services-based applications we have less insight to and control over the performance needs of those applications. No longer can we easily benchmark an application and then apply requisite network topologies to satisfy its typical performance needs. Nor can we simply continue to “throw bandwidth” at performance problems. Next generation software-defined networks (SDN) must intelligently adapt topologies and quality of service based on numerous external factors driven by and collected from a variety of loosely-coupled, federated sources. The key to these externalized applications is the ability of the control plane to maintain a holistic, coherent view of the state of network, which can then be used to construct new demand-sensitive topologies, routes and services. Just like we’ve seen with servers and storage, the network view will consist of virtualized resources and routes that will be abstracted from the underlying physical devices.
Next generation intelligent networks will not be driven just from an application and workload placement view. The user experience will be a critical part of the equation, with BYOD providing the catalyst to create network intelligence that will enable a consistent and continuous user experience as we move from office to home to public locations. The intelligent enterprise network will have to provide a seamless transition of the work experience through new approaches such as the implementation of distributed networks of small cell base stations that not only bridge the gap between internal and external networks, but also provide state-maintaining location, session and experience awareness.
In and of itself, the enterprise network will be very data intensive in terms of state awareness and management. The explosion of machine-to-machine communications is going to require even more levels of device abstraction, virtualization and state management as millions upon millions of short-term communications channels are initiated and terminated to support instantaneous bursts of contextual state and awareness information. The intelligent network will not only need to scale to meet the massive influx of data, it must also understand how to communicate with an ever-increasing variety of devices.
The key to intelligent networks will be their ability externalize and virtualize the view and control of the underlying resources. Significant progress in defining standards and protocols is being made by organizations such as ONF and the Network Functions Virtualisation Group. Good work, but we’re still a ways from replacing bespoke network boxes with merchant silicon-based devices that can be “software repurposed” to meet different workload, application and user experience requirements. Stay tuned though – it’s getting exciting!
