For some, it is a predetermined conclusion that white box switching is the future. They envision a world of commodity switches and independently distributed software. Some IT shops will manage their own Linux distributions. Others will use commercial offerings, leveraging a support model that comes with it. Regardless of whether this describes your particular future, we ought to take a moment and ask ourselves how we got here.
The prevailing thought is that the reason white box switching if a fait accompli is that the economics just make sense. Why pay for sophisticated hardware when most of the magic is in the software? Why incur addition expense for a handful of esoteric features that you could probably design around anyway?
There are a couple of issues with this rationale. First, while switch and router pricing certainly skews towards hardware, vendor R&D budgets are heavily biased towards software. The hardware has such high margins largely because virtually all of the pricing is aimed that way. For the large vendors, somewhere north of 80% of their R&D spend is geared for software.
But the bigger issue in my mind is if this is all a CapEx avoidance game, why is it taking off now? Prices have been high for years. Sure, capacity is growing exponentially, so adding bandwidth every year gets harder and harder. But have we hit some tipping point suddenly on the capacity side? Meanwhile, switch pricing has come down steadily over the years, driven partly by merchant silicon and mostly by competition. There is no greater enemy of lofty product margins than healthy competition. Arista’s continued share gains are enough to bring pricing down for everyone, and the fact that SDN has brought a bunch of new companies into the fray can only help keep things moving in the right direction.
And even if white box switching was primarily a CapEx thing, only those companies that deploy switches en masse would stand to save any sizable investment. Saving several thousand dollars is great. Multiplied over hundreds or thousands of switches, the difference is meaningful. Spread over a half dozen switches though? Savings is always good, but I don’t see the compelling event there.
None of this is to say that white box switching is not good. Rather I am suggesting that it’s rise is rooted in something other than profit margins and CapEx budgets.
When an industry steps up and says that they would rather build and maintain their own products than buy prefabricated stuff, it is more than just firing a shot across the margin bow. It’s an organic response to solutions that have collectively fallen short. It’s an indictment not of a single company but of an entire industry. It signals to all of us just how far short the networking industry is falling.
The question vendors need to ask is where have they been falling short?
Based on industry chatter, it is tempting to think that it’s all about price. It is not surprising to see a rash of moves all designed to lower the price per port. A bigger box here, a collapsed box there… drop the price. A partnership here, merchant silicon there… drop the price. But while price is important, it is really just the manifestation of years of growing discontent. The issue isn’t price, but margins have become symbolic of how the industry has under delivered. Price must fall.
But as it falls, more has to be done. Obviously SDN, NFV, network virtualization, and DevOps all aim to drive long-term OpEx down through workflow automation and workload delegation. Software is the lynchpin for all of these efforts, and so the DIY movement is correctly focused there. The question is not whether software is the future, but really whether the software component vendors can out-OpEx the more established vendors. To be fair, the likes of Cumulus don’t actually have to completely outperform their legacy counterparts; because of the growing tide of discontent, they just need to reach some level of parity. This should be frightening for some folks.
So what are the implications?
For people who have been caught up in all the CapEx talk, things could get ugly. Incorrectly diagnosing the problems means that roadmaps will not intersect the correct long-term opportunities. CapEx is a giant red herring, and companies that rotate their entire strategy around CapEx will find themselves ill-prepared for a software future that revolves more around OpEx in a highly-automated, more fully-integrated IT universe.
And if the end game is not only lower price (certainly, some of it is price – I don’t mean to discount that argument entirely) but also integration and orchestration, then there will be avenues to compete. But the features here are different than they have been for decades. Companies tooled for performance improvements and BGP knobs will find that their engineering resources do not match their customers’ long-term goals.
Worse yet, every customer win from one of those BGP features simply provides temporary comfort. It extends the life of the existing network by some small amount of time, which allows the customer to line up the pieces necessary to make their own transition. When that transition time comes, will the customer wait for the vendor to catch up? Not a chance. The environment that has led to the DIY movement is particularly unforgiving, especially for the vendors that have been complicit in its rise.
Whatever the outcome, I think we all ought to be somewhat aware of not only where we are but why we are here. White box or not, it’s instructive about how to compete going forward.[Today’s fun fact: The plastic things on the end of shoelaces are called aglets. I wonder why Texas A&M named their mascot after shoelaces?]