Victoria - Fotolia
Packet Pushers Interactive
Published: 17 Nov 2015
The jury's still out on software-defined networking, but white box switching just might stand the test of time.
Disruption is a technology buzzword applied to any new idea that threatens to overturn the status quo, and the field of networking has seen a great deal of disruption over the last five years. Specifically, software-defined networking (SDN) resurrected the idea of centralized control, where applications manage network configuration, issuing commands to the network hardware and software within its domain via a centralized controller.
White box switching is a byproduct of SDN, decoupling the physical network switch hardware from its network operating system (NOS) software. While vendors and practitioners alike are still pondering the merits of SDN and centralized network control, white box switching might survive on its own merits--even if SDN falls by the wayside.
Also known as open networking, the idea behind white box switching is to allow an organization to buy a switch of its choosing and pair it with its preferred NOS. While IT practitioners are used to this sort of flexibility with servers, networking is another matter, as vendors traditionally couple their software tightly with their own hardware.
The coupled approach allows each vendor to write its NOS in ways that fully exploit the hardware. It also means that a vendor can charge a premium. If an organization has standardized on specific processes to configure its switches, it is tied to a specific command-line interface (CLI) to execute those processes. This expensive state of affairs is bothersome to savvy IT organizations that feel they are being overcharged for the hardware to get the software they want.
Making the financial sting a little worse, many network switches contain largely the same hardware as their white box counterparts, based on merchant silicon. That is to say, the networking vendor outsourced the chip design and manufacturing to a third party such as Broadcom (under agreement to be acquired by Avago Technologies), which has Trident II, Trident II Plus and Tomahawk chipsets inside proprietary and white box switches alike -- with identical performance numbers.
The question then becomes one of differentiation. Why pay a premium for a network switch with internals that are the same as several competing switches? The differentiator has always been the network operating system. By decoupling network hardware and NOS, white box switching allows companies to evaluate hardware on its capabilities (and costs) and pair that with a NOS that competes on its own merits.
What white box networking changes
Is white box networking really a good idea? Let's consider when white box switches make sense, and the implications of this networking model. The open networking supplier market is quite young, and the technology won't be right for everyone.
Perhaps the biggest argument for white box switching is choice. It gives organizations the ability to change a NOS at will, without buying new switch hardware. This freedom affords a business more flexibility in how it conducts its network operations.
Making an operational change is difficult. SDN, however, has driven network vendors to add a feature in their switches called network programmability, using application programming interfaces (APIs) and other standards-based protocols like OpenFlow or NETCONF. This means that network operators don't have to interact with the command line to configure a switch. Rather, other tools, such as pre-made scripts or third party applications can be used. This reduces the dependence on specific CLI syntax.
The big catch here is that network programmability doesn't guarantee that interacting with APIs, OpenFlow or NETCONF will be the same experience on every network switch. In fact, the experience can vary widely. In that sense, the industry hasn't moved the ball far enough ahead to make network configuration a generic, platform-agnostic task. While the advantage of a NOS choice is a significant one, there is still an operational effect felt in a NOS change.
The cost factor of opening up
In general, an organization will find choice a compelling argument for white box networking -- if it configures its network switches using an SDN application provided by a vendor. That way, the IT organization doesn't have to worry about how the switches are configured, and IT staff can interact with the application that handles the configuration details. The organization should also have in-house developers ready to write custom networking applications as the business defines new requirements.
The other major driver for white box network switching is cost. White box switches make sense for organizations that clearly understand the total cost of ownership of their existing network infrastructure. Only then can an accurate cost comparison be made to the total cost of ownership (TCO) of proprietary and white box switching.
How to investigate white box switching
White box switching could be asking a great deal of your organization's IT operations team. Here are some key points to consider:
- Understand why you need it. The white box model can be a powerful enabler, but only if your existing networking vendor harms your business agility or budget. This is not the sort of change to be undertaken for sheer novelty. Make sure the drivers are there.
- Know your operational model. Having a deep understanding of your organization's network operations is critical when evaluating white box switching. The change will affect staff, service delivery and system monitoring, and knowing the existing operational model will help you plan a path to the new model.
- Start small and test carefully. White box networking is not an all-or-nothing proposition. For organizations with existing infrastructure, a simple place to start is in a lab, to validate switch hardware and software capabilities, provisioning processes, management capabilities and integration with the existing network. Only then should a few switches be deployed into production, in a low-risk area.
Intuitively, a white box hardware switch should be cheaper than a premium branded switch with the same internals. This provides a potential benefit in cost of acquisition. The NOS to run on the hardware will vary in price and purchasing model.
When thinking through costs, consider that white box TCO is not merely the sum of a switch and a NOS. Some NOS offerings require a subscription plan. Other potential costs include support contracts, staff training and integration with the legacy network.
In the end, the idea of decoupling network hardware from its software makes sense for many of the same reasons it made sense in the server world. However, the future of white box switching is far from assured, with big networking vendors naturally reticent to embrace an open model that makes frees customers to make more relationship choices.
To learn about vendors that embrace -- or reject white box switching, click here for part two.