Skip to main content

Contents

DriveNets #NFD22

Contents
DriveNets logo

Recently I attended Networking Field Day 22 (NFD22) in San Jose, CA as a delegate. By attending NFD you may be exposed to companies that you may not normally come across. The first company that presented, DriveNets, was one of those companies and here is their story.

DriveNets was founded a little over 4 years ago with a mission to “disrupt network economics with growth-ready cloud-based infrastructure.” Their main target market is the Tier 1 ISP ( ATT, NTT, Level3, etc). They started with four things in mind.

  1. Software-Based
  2. Scale
  3. Open Architecture
  4. Simple Operation

When you look at a “legacy service provider” network, you will see that there are four primary layers. You have

Slide titled Service Providers Complexity showing a legacy network architecture with dedicated hardware at each layer

As can be seen in the diagram above, each of these layers has its own dedicated hardware. As you could imagine, the cost involved with scaling this type of design out can be costly and don’t forget that managing larger and larger networks bring in their own challenges.

Slide titled Network Cloud: Three Levels of Disaggregation, from hardware and software through router architecture to cloud-native software

The goal of DriveNets is to help enable Service Providers (SP) to move to a more Network Cloud approach. This means disaggregating the traditional router into building blocks consisting of Routing Engine, Fabric, and Packet Forwarders (white boxes). By doing this, the SP gains scale at a greater rate as they can just keep adding packet forwarders to the network to keep up with port demand, growth, and increases in port speeds.

Slide titled Disaggregating the Traditional Router, comparing a monolithic chassis with a disaggregated router built from servers and white boxes

The above diagram shows a traditional chassis-based solution next to the DriveNets solution. The chassis can be seen as a monolithic device that is all contained within a single piece of sheet metal. Compared to what DriveNets is offering where you use x86 based servers for your controller and white-boxes for your fabric (NCF) and packet forwarders (NCP). If you need to scale at any level, you can built-out as necessary to support the demand.

DriveNets Network Cloud operational model slide comparing a legacy network architecture with a cloud-based one across the last mile, metro, edge and core

So what would it look like at an SP that was using DriveNets for their network? Well, instead of having large quantities of managed devices at each layer, they could have a single box for each layer – a box that scales out as the business grows. As they need more ports, they just add a new packet forwarder to the network and connect it to the existing fabric and controller.

Slide titled Abstraction and Orchestration of the Router Elements, showing DNOS containers, a base OS and an orchestration agent on each platform

DriveNet has their own Linux based OS called DNOS. This is a natively distributed NOS that makes extensive use of Docker containers. This allows them to deploy what they test and develop on as once the container is built, they can move it to production and know it will work the same. This is also interesting as it allows you to upgrade containers individually and not have to upgrade the entire NOS when a new feature is added.

Slide titled A New Economic Model showing a fixed software license on white box hardware against legacy network cost as capacity grows over five years

A real differentiator with DriveNets for SPs is their economic model. What they have decided to do is sell their software license at a FIXED price. Not on a per-port, per box, etc – but for a fixed price for the controller deployment. So as that network grows, the cost savings increase – really interesting.

I think what DriveNets is offering is really interesting for Service Providers. It offers them a way to have an SDN function without having to have in-house programmers to manage and maintain the solution. It also allows the SP to grow their network on an almost per-port basis, and at a potentially lower cost ( white box vs vendor cost).

I do feel that there is a risk with this type of deployment though, as you scale the port count out – your blast radius for problems increases. When you are running a chassis solution that has ~260 10G ports on it, a single chassis out of 5 having problems is a minor headache. With the DriveNets, your virtual router with over 7000 10G ports has a problem – you might be in for a bad day. Granted this could be mitigated by design and building for failure – but it is a thought that we need to keep in mind.

Be sure to check out the DriveNets presentation videos below. There is so much there that it is hard to capture in a single post.


You can check out all the videos from Network Field Day #22 over at Tech Field Day’s YouTube Channel.


Disclaimer: I was invited to Networking Field Day 22 with GestaltIT covering travel and accommodation. There is no requirement to blog, promote, or produce any content from the event. This post is my opinion and my opinion alone.