The network operating system (NOS) is taking on greater significance both in the data center and on carrier-level networks as proprietary platforms are replaced with white box infrastructure.
The need for increased flexibility and fully virtualized network stacks all but mandates a clean break between hardware and software as data users become less tolerant of downtime, crimped performance and anything else that interferes with their access to data and services.
Just like the PC operating system, the NOS is essentially the glue that binds everything, hardware and software, together. In its ideal form, the operating system works quietly behind the scenes, coordinating resources, managing data flows and performing system checks. The main different between the PC OS and the NOS, of course, is that the NOS is geared toward switches, routers and controllers rather than disk drives and processors.
With networking performing such a complex, and highly critical, function over the years, it’s not surprising that the NOS became hardwired into the leading manufacturers’ platforms. In the last decade or so, Cisco, Juniper and numerous other network companies have all opened up their operating systems to some degree, albeit optimized for their own hardware.
The rise of commodity white box hardware, however, has opened up the opportunity to experiment with new operating systems from a wide range of vendors, some with more networking experience than others. For organizations contemplating such a change, then, what features and capabilities should the OS have? And does it make sense to go with a pure-play open solution or a vendor system that may define “open” as anything between a RESTful API to something built on top of Linux source code?
According to market analyst Lee Doyle, the answer to the second question depends largely on the balance between integrating new network infrastructure with legacy plant versus implementing the kind of flexibility needed to craft best-of-breed solutions. White box solutions are certainly cheaper than proprietary, sometimes by as much as half, but few organizations have the expertise to install the OS on their own. This means a trusted VAR or systems integrator needs to be brought in. While this adds to the overall cost, it also provides an opportunity to create highly customized network solutions for individual workloads and business models.
This works best, of course, if the boxes are deployed on greenfield infrastructure. Integrating disparate systems on the OS level is tricky at best and could lead to significant performance problems at worst.
Once the decision to go with an independent NOS has been made, what considerations should go into selecting a platform? Perhaps the most critical attribute is interoperability. By nature, networking depends on careful coordination of multiple elements, so any miscommunication between one set of resources with another introduces latency, if not outright service disruption. This interoperability should extend not just to network hardware but to management systems, security software and the many development platforms that organizations will use to streamline and optimize virtual network infrastructure in the field.
On the carrier level, of course, interoperability is crucial given the need to scale network resources to extreme degrees. This will invariably involve connecting to numerous third-party devices, which must then be utilized for highly segmented routing architectures and functions like telemetry streaming and Layer 3 MPLS services.
But probably the biggest contribution that the NOS will make in the new era of networking is the ability to lift entire architectures off of hardware and deploy them as virtual fabrics. In addition to the cost savings and operational benefits, this will make data communications more resilient, secure and reliable than anything that has come before.
In an age where data is king, the NOS becomes the primary means to deliver the most vital service of all: preservation of the networks that keep the world in touch with the digital ecosystem.