Big Switch Networks, the startup that was once the face of software-defined networking (SDN) to many, has had it rough lately.

Overlay networks have become the rage, rather than the OpenFlow protocol that brought Big Switch and SDN to fame. The creation of the OpenDaylight Project in April threatened to marginalize OpenFlow controllers as products — and was followed by the blow of having OpenDaylight bypass Big Switch's controller in favor of a Cisco/IBM proposal.

Add a couple of executive departures and vague rumors of layoffs, and — yeah, it adds up to a rough year.

Big Switch has decided it's time to explain its position. In media interviews this week, the company is saying it's scrapped the overlay version of its Big Virtual Switch, in favor of a new foundation that challenges overlays altogether.

And amid the turmoil, the company claims it's found a gusher of a revenue stream. "We're going to have our highest-revenue quarter by quite a bit, but it's still been tough on our people," says Kyle Forster, a company founder who now also carries the title of president.

The problem Big Switch found with overlays was the Erector-set effect: lots of loose pieces that took lots of skill to assemble. Moreover, Big Switch had to pair up with equipment and software providers to make it work; the matrix of partnerships "became very complicated very fast," Forster says.

Big Switch also found the overlay model impractical to automate, one reason being that the overlay depended on the states of routers, firewalls, and other elements.

So, Big Virtual Switch turned away from overlays in the first quarter of this year. The company's new plan for network virtualization involves using one controller to provision physical and virtual switches in tandem — a cloud fabric that the company informally calls Cloud P+V.

The physical-switch side would be covered by white-box switches that would automatically download Switch Light software from Big Switch's controller, similar to the Pre-Execution Environment (PXE) booting that servers can do.

"It's simpler, and that's what's it's going to take to get past the 1-percenters," Forster says, referring to the most advanced 1 percent of the customer base.

It also could elevate Big Switch's status as the single vendor to support this part of the network. One knock on white-box switches is that there's nowhere to go for help; by being the sole source of software, Big Switch would become the support desk for issues that aren't obviously hardware problems.

The new cloud fabric is based on OpenFlow, with a few extensions Big Switch had to write. Creating OpenFlow extensions has been routine for the company, Forster says; Big Switch tries to stay 18 months ahead of the standard, and it makes the extensions available in the open-source Switch Light and Floodlight software, to deflect any claims that it's creating a proprietary OpenFlow spur.

By proposing an overlay alternative, Big Switch is going against the grain; only Cisco's Insieme, with the Application-Centric Infrastructure concept, is chasing a similarly non-overlay architecture, Forster believes.

The shift away from overlays did require personnel changes, Forster says.

"In overlays, you don't need to have that much networking background. The networking principles are far less important," he says. But a Cloud P+V plan needs equal parts virtualization and networking talent. That's meant cutting some people while also hiring; Forster claims the two have balanced out but doesn't have an exact headcount.

That shift in priorities might explain the departures of executives Andrew Harding and Howie Xu (who came from a virtualization heritage at VMware) in recent months. And there's one more to add: Jason Matlof, VP of marketing, recently left to join A10 Networks.

Until the cloud fabric can get off the ground, Big Switch's flagship product will be its monitoring fabric.

Conceptually, it's similar to the network packet brokers (NPBs) offered by companies such as Gigamon and VSS Monitoring: a switched mini-network that connects network taps to test and analysis equipment. The difference is that NPBs tend to ship into fix-it situations, where an operator has amassed lots of these things and needs to connect and organize them.

By contrast, Big Switch's monitoring fabric targets huge, organized situations, where network monitoring is being carefully knitted into a data-center's blueprint, and on a larger scale — hundreds of network taps rather than dozens. Forster cites one customer that has 16 teams monitoring the data center, each with its own farm of tools.

It's very much a 1-percenter product, Forster says. And yet, it's got a lot of pent-up demand, because this kind of fabric is hard to do. For example, multi-tenancy issues tend to crop up, such as having to manage the overlapping priorities of different groups using the fabric. "There are some very hard math problems to solve there. There's some Ph.D. work under there," Forster says.

With SDN being an uncertain bet two or three years ago, Big Switch targeted those high-end customers intentionally, but there was a price involved. "If you're going to play that game, going after the super high end, that means you have to be patient with your revenue," Forster says.

Big Switch has opted to keep that patience, putting the cloud fabric through a beta test intended to last 12 to 18 months. The company isn't saying how long ago that cycle lasted.

The long beta period is part of Big Switch's culture, a way to groom a product into production worthiness, Forster says. It takes money, but Big Switch claims to have that. Forster dismisses a rumor that says the company has $10 million left in the bank; it's more than twice that, and Big Switch hasn't yet tapped the funds from its B round ($25 million raised in October), he says.

Many of the beta customers are among that high-end, 1-percenter range, but about one-third to half are not, reflecting the cloud product's broader appeal compared to the monitoring fabric, Forster says.