This post is in reply of Gianpaolo’s post “White Boxes for Everyone?” appeared on his blog last November and recently shared by Ivan Pepelnjak.
First, let me state that I don’t disagree with him nor with Ivan, but I also believe the matter is worth some more in-depth analysis.
His thesis is that the whole white-box movement makes sense for the FANGs of the world, as their scale and their capability of attract skillset allows them to develop, operate and maintain custom-built software at scale. Smart-people, dumb (and cheap) devices.
That is absolutely correct and spot-on.
He then compares the FANGs to the common Enterprise company, where the lack of skills demands for an inverse approach: dumb people, smart devices (and smart vendors).
I worked on bright and white-box switching for a while now and what I can say is that you don’t necessarily need the smartest of people to run a bright-box architecture. You don’t need to have developers to tailor your own custom NOS to make sense out of network disaggregation.
Disaggregation is, first and foremost, about enabling faster innovation cycles, removing dependencies that slowed down the introduction of new and more efficient HW and SW. Or enable easier and less dramatic changes of strategy when the SW or HW vendors needs to be substituted.
I don’t think network disaggregation is a black-or-white game. There can be multiple intermediate steps and nuances along the way of disaggregation.
A few examples:
Say you are an Arista customer (just for the sake of the argument): one day you want to experiment bright-box switching, and run vEOS on top of a general-purpose Broadcom-chipset based switch. You can. (doesn’t mean you should)
You want to run it on an HPE/Juniper/Dell bright-box because you can leverage the traditional vendor-led support model? You can. Most of the bright-box players offer support for the software layer as well through cross-vendors agreements, if you really need to have a single-point-of-contact (i.e. a single TAC to talk to).
It’s not only the capex savings, it’s not only the faster innovation, but it’s also about the culture.
As Gian Paolo correctly points out, vendors are trying to dumb network engineers down more and more, as they realised this is the only way they can replace the old lock-in by vertical integration with a different one, a cultural one.
The older generations of network engineers learnt networking the hard way, by making mistakes, looking at CLI output, etc.
Vendors seem to not want us to do that anymore. There’s nothing wrong about helping everybody doing their jobs quicker and more efficiently, but there is absolutely something wrong when complexity is completely hidden from the users/engineers, as they’ll have no drive to learn how things really work.
Bright-box and white-box I believe enable a more mature culture. I call it the culture of “doing”. Having the possibility of running a Linux operating system which is (in most of the cases) open, allows engineers to come up with new ideas to do things differently, on the network device itself, interacting with more open platforms, APIs, data-models, leveraging well-known interfaces and OSs.
Having said that, I’m not saying everyone should jump on the bandwagon (in technology we all tend to do that a bit too often), and look at “disaggregation-white-bright-box” blindfolded. We always shall start from our business requirements, not only in terms of network functionalities and operating models, but also taking the requirements of changing the mindset in our companies’ workforce into account.
It’s true. Everyone is dumb and – even worse – we are getting dumber. I believe that disaggregation is a lighthouse that show us the way to a world where we don’t HAVE to become dumber, if we don’t want to, and a world of doers and not one of “followers”.