Best-of-breed networking: There’s an opportunity for white boxes in the WAN

Everyone wants the best network, so they say, but as usual with “bestness” goals, there’s not much agreement on how to achieve it. Do you go for best-of-breed in your equipment, or maybe for three vendors per product area, or maybe open white-box networks? A bit over three-quarters of enterprises said in 2020 that open networks and white boxes would give them best of breed options, but what’s happened since then seems like a blast from the past, vendor-wise.

Let’s start with an interesting truth. Well over 95% of established enterprises who can look back to the origins of their networks say that they started off with a single-vendor network. Enterprises who launched within the last five years (which are very few) account for almost all those who didn’t start off with one vendor. If you think about it, this is logical, because starting off with networking is daunting enough without adding in the complication of network integration and the management of multiple sources of technology.

Truth number two is also interesting. Today, 37% of those same enterprises say they have multivendor networks and another 9% say they have open networks. In the net, a bit over 40% of those single-vendor enterprises have shifted their vendor mix over time, and the biggest reason given is pedestrian: Acquisitions and mergers brought in other vendors. The second-biggest reason is more interesting: Their single vendor did something dramatic to discredit themselves.

Enterprises say that network vendors often disappoint them, but rarely enough to drop them.  In the last 10 years, I identified only three cases where an enterprise’s single vendor messed up badly enough to be well and fully dumped. Why? Replacing a lot of gear is a major budgeting problem. Not only do you have to buy new gear, you have to accommodate the write-down of existing equipment too. Over 90% of network budgets in 2023 will be directed at simply maintaining current application and information access, so any changes in the network will be subject to stringent cost controls. The keep-the-lights-on focus also discourages introducing new vendors who may pose new integration and network-stability risks.

All this offers the primary vendor major advantages, since the vendor with the biggest piece of the pie can usually afford to commit more resources to the customer, and offer bigger discounts. Over the last five years, the primary vendors won three-quarters of the deals, according to enterprises. Unless this primary vendor creates a new problem, they seem to gradually gain strength even after a they’ve created a major problem that cost them some of an enterprise’s business, and to recover their dominance over a period of four or five years.

Dominance, but not necessarily exclusivity. In fact, even single-vendor networks aren’t always singular. Digging into even enterprises that claim to be single-vendor shows that “single-vendor” enterprises often have more than one vendor after all. Over 30% who claim to be single-vendor turn out to have a second vendor in play.  The reason is that, for enterprises, there’s networks and then there’s networks.

Most CIOs and the network professionals on their staff tend to think of “the network” as being “the WAN”.  Data-center networks, the LAN switches that tie all the servers together, are regularly forgotten when talking about their network vendors. That’s because they tend to evolve because of IT changes—not changes in “network” requirements. That’s important, because the data-center network is also where enterprises think white boxes and open networks are most credible. Well over a third of enterprises are thinking of those technologies in the data center, compared to less than 10% thinking of them in the WAN.

Data center switching opened up because new requirements drove an opportunity to modernize. The explosion in virtualization and horizontal traffic between components of applications created a need for significant capacity expansion and an improvement in traffic routing within the data center. Things like software-defined networking (SDN) and centralized route control are very practical within the data center, and there are really only two kinds of interfaces to worry about there—the ones to the servers and an exit interface to the company VPN, the internet, or both. There was also a lot of publicity given to white-box data centers, and plenty of vendors interested in offering them, and this combined with expansion funding to create the opportunity for something new and different.

Why then aren’t enterprises moving wholesale to white boxes in the data center? It’s difficult to justify a big change in networks that aren’t subject to major requirements changes. Data-center networks evolve and are modernized, which means that adoption of white boxes is still limited to places where additional capacity is needed. In fact, the majority of enterprises that have white-box switches in the data center have them in parts of the data center that were recently added or underwent major changes, usually due to virtual-hosting requirements. It’s this same evolution and modernization that could support the spread of white-box open-network technology to the WAN.

Could because no, this would not be a slam dunk. Enterprises, even those with white-box data-center LAN switches in use, think that white-box WANs require more integration effort. But they also believe they’re cheaper, and enterprises that have deployed them in the data center say that’s true. All this suggests that enterprises could be won over…but there’s that could qualifier again. Only 9% of enterprises say that a credible vendor has offered them a white-box router, and two thirds say they don’t know a credible white-box open-model vendor at all.

That may be the lesson here, one that takes us back to those first network commitments. Enterprises don’t like network integration. They start single-vendor, they drift into something more open, and they drift back. There’s no real advocacy for an open enterprise network option, no organization to center stories around, no practice guides from credible sources—none of the things that make CIOs, and more importantly CEOs and CFOs, comfortable. And unlike operators, whose capex is focused mostly on network equipment, the enterprise network is usually a small piece of enterprise investments.

And none of this may change. There is price pressure on enterprise networks, to be sure, but so far the major vendors are discounting enough to keep things under control. Absent some sudden radical shift for enterprise networks—as 5G was for operator networks—vendors may have time to adapt to the gradual evolution and modernization we’re seeing in data-center networks. Evangelizing white boxes to an uneducated enterprise market is a lot different from calling on a dozen Tier One carriers. And in the enterprise, discounts work wonders staving off competition.

In any event, enterprise networks both in and out of the data center face other challenges, not the least being the acceleration of virtual networking. VPNs eliminated a lot of routers. SD-WAN and cloud-hosted SASE could eliminate even more, and what’s left will be increasingly tied to those exit points in the data center. If open-network trends continue there, then they could gradually move outward; some SD-WANs are already hosted on white boxes, after all. If this happens, with a truly open model for networking possible, we might finally see best-of-breed networks become a reality.