Networking startup Volta Networks -- winner of TechTarget's Network Innovation Award -- is one of a handful of young companies aiming to change the routing game. Drawing inspiration from server virtualization and software-defined networking, or SDN, Volta created a disaggregated, cloud-based routing platform that allows service providers to run up to 255 virtual routing instances on a single white box switch. Decoupling software from hardware and moving the control plane to the cloud can lower costs and increase agility, which is of growing interest at the edge as the 5G era dawns.
In this Q&A, Hugh Kelly, Volta's vice president of marketing and a network industry veteran, discusses the networking startup's cloud-based routing technology.
Editor's note: This interview has been lightly edited for length and clarity.
What need did Volta Networks set out to meet?
Hugh Kelly: Over the past 10 years or so, there's been a very strong interest in SDN. Initially that meant using OpenFlow as a centralized control point for all of a network's routers, but that never really took off. I think it met a lot of resistance from the routing manufacturers that were concerned about how well it could scale in large provider networks. But the basic concept was one that everyone became excited about -- separating software and hardware to give the customer more options, better control over their costs and the ability to optimize for their particular goals.
Hugh Kelly Vice president of marketing, Volta Networks
We embraced this disaggregated model, and our goal was to scale the number of virtual routers that can run on a single piece of equipment -- a white box switch, for example. We wanted to bring the flexibility of SDN to the routing market, but without trying to put more and more CPU on white boxes, which raises their price and undercuts cost savings. Instead, we realized the easiest and least expensive way to scale processing and run multiple virtual routers on a single low-cost device would be to move the control plane to the cloud. We didn't have to introduce new protocols or change how customers built their networks -- we simply changed where the control plane ran.
As a side note, when we say control plane, we're talking about Junos software from Juniper, IOS from Cisco and all the routing protocols and associated management protocols -- [Border Gateway Protocol], [Open Shortest Path First], [Intermediate System-to-Intermediate System], MPLS and segment routing. All of those things are very processing intensive.
We've also embraced Free Range Routing and are active participants and contributors to the open source model and community. For those who aren't familiar with Free Range Routing, it is a recent iteration of a set of open source routing protocols, which has robust support from a number of vendors -- folks who are saying, 'no point in each reinventing the wheel.' Open source communities either take on a life of their own or they don't, and Free Range Routing and networking clearly has.
Can you explain how Volta Networks' cloud-based routing technology works?
Kelly: When we talk about a virtual router, what we really mean is a dedicated instance -- essentially a container and a cloud-native application. Each virtual router has its own protocol and management interface and is really set up as a complete administrative domain, and you can have multiple of these routing actions associated with a single switch. We just dedicate resources -- meaning ports on a switch -- to that virtual router. So virtual routers run in the cloud and are logically associated with certain ports.
The question customers always ask is, 'How well does it work if the control plane is separated from the underlying hardware?' And the answer is, if you build a good solid network between them, we can throw more CPUs at the control plane and actually get it to perform better. That's easy to do in the cloud -- you can scale that processing and memory much more easily. We use a secure interface between the boxes and have optimized communication between them, so it's incredibly efficient.
Then people say, 'All right, so each of these routing protocols creates its own forwarding and routing information, with a table that has to be downloaded. How do you convert that to the forwarding information base, or FIB, that's physically resident on the box?' We have some patent pending technology that allows us to aggregate, compress and manage that information, transparent to the customer.
We've also embraced a service model, so everything's very API-driven. Even though we support command-line interface -- like every Cisco or Juniper network engineer would expect -- we are very early adopters of API, NETCONF and YANG service models. We view those as critical for automating large-scale service provider networks, which, in turn, are essential in generating new enterprise services.
How does Volta Networks envision the landscape changing in 2020 and beyond?
Kelly: Service providers' recent embrace of open networking has been really noteworthy. Providers were originally reluctant to use white boxes because the technical capabilities of ASICs [applications-specific integrated circuits] like Tomahawk and Trident weren't really suited to large-scale routing and the big routing tables that a service provider needs. But now -- with Broadcom's Qumran and Jericho chips -- platforms have substantially more capacity, and vendors like Edgecore have been building boxes for the carrier market.
In the past year or so, service providers went from having just a couple of white boxes in the lab to being really aggressive with testing. We're seeing them really embrace what is possible with these commercial, commodity-off-the-shelf ASICs, and we think we're at something of a tipping point. We expect adoption to really take off in the next year. That, in turn, will make it feasible for providers to drive all sorts of new enterprise services, such as IoT, augmented reality, virtual reality, smart cities and connected cars.
Because of all of this and the enormous potential providers see in 5G, we expect to see routing needs at the edge evolve over the long term. Providers are really focused on getting their investments to a critical mass, so they can start to realize some of these new applications. All of this requires more routing, which is obviously easier and lower cost if you're using virtual and cloud-based routing, rather than something proprietary and hardware based.