DriveNets' disaggregated router tech wins innovation award

TechTarget has named DriveNets a winner of its Network Innovation Award in recognition of the company's software-based routing technology. The networking startup -- based in Ra'anana, Israel -- says its disaggregated router model will help carriers scale their networks quickly and cost-effectively in the tradition of hyperscale data centers. DriveNets' software can run on a single white box or across a cluster of hundreds, offering providers immense flexibility to meet demand as it grows.

SearchNetworking spoke with Inbar Lasser-Raab, DriveNets' chief marketing officer, to learn more.

Editor's note: This interview has been lightly edited for length and clarity.

Why did DriveNets decide to focus on a distributed, disaggregated router model?

Inbar Lasser-Raab: Our founders kept hearing again and again from service providers that they were really looking to scale their networks to support the cloud era and new applications -- like artificial intelligence processing, virtual reality and streaming services -- while also maintaining profitability. Networks are growing roughly 50% a year, and as networks grow, revenue falls faster than operating costs do. On one hand, there is this need to scale the network to new heights, but on the other hand, profitability is dropping. How do you bridge that?

Hyperscalers faced a similar challenge in cloud. As more and more applications were deployed, revenue didn't rise as fast as the need for bigger and bigger infrastructure. Their answer was to disaggregate software and hardware, develop software to run on standard white boxes and add white boxes as they scaled.

It can be a single box, four boxes, seven boxes, 50 boxes or 200 boxes -- it's still a single router. It scales as you need it to.

So, we said, 'Let's solve this the same way they did, building software that can run huge core and edge networks.' The hardware comes from multiple original design manufacturers [ODMs], designed to align with AT&T's distributed, disaggregated chassis model. The ODMs buy the software -- which is the router itself -- from us. It can run on a single white box at 4 terabits per second or in containers across 200 white boxes supporting 200 terabits per second. It can be a single box, four boxes, seven boxes, 50 boxes or 200 boxes -- it's still a single router. It scales as you need it to.

Who is the target customer for DriveNets' disaggregated router software?

Lasser-Raab: We're looking at Tier 1 and Tier 2 operators that are building hyperscale networks. They may start small, augmenting the network they have. But over time -- since capacity doubles every two to three years or so -- the new network built on white boxes will be larger than the original.

How receptive are carriers to the idea of a disaggregated router model, given the significant changes it would mean for their infrastructure?

Lasser-Raab: We spent all of last year educating the market on the need for the disaggregated model, and some operators picked up on it really quickly. Service providers today have multiple router models in their networks that they have to maintain, and every time they reach capacity, they have to decide, 'Do I do a forklift upgrade and bring in a larger router, or do I put another router next to my existing one and waste bandwidth connecting them?' They always have this challenge. We say, 'Look at how fast hyperscale clouds have grown and how they've done it: They just bring in more white boxes.'

I think service providers are now buying into disaggregation as a concept. This year, it's about showing them why we do it well. A lot of people talk about the concept of disaggregated routing, but we actually have technology that works. When we do proof-of-concept demonstrations, service providers see that. We have a lab downstairs in the DriveNets building that has a cluster of 192 terabits running as a single router. It always wows customers when they visit.

What other benefits does DriveNets' disaggregated router architecture have?

Lasser-Raab: Think about cloud today: You can decide you want to run your application on a certain server location, or you can go serverless and have your application just run everywhere. We, in a way, do the same. We look at the network as a pool of ports that is a fully shared resource, and there are tremendous efficiencies built into that.

The traditional network is really underutilized because you build for peak use in every location -- access, edge, core. We actually treat it, on a single-location basis, as a pool of ports. It's a fully virtualized concept from a technology perspective. We take the big router and break it into building blocks and run a cloud-based software on top of it. It takes people a while to really grasp the full benefits of the model, but the reaction is ultimately, 'Wow, how come nobody thought of that?'