Vendors have long promised a self-driving network that can do all the under-the-hood tuning and pruning without human assistance. But what enterprises seem to truly want is a network that is low-tech, and low-touch, like the cloud.
Hedgehog, a networking startup, has a solution that they might be looking for.
Low-Tech Equals Low-Effort
Digital environments have grown infinitely more complex. The solutions that have landed on top of them have only made the job harder for operators.
In practice, the myriad principles and solutions aimed at making networking processes low-effort all feel like different paths leading to the same thing – a growing operational burden.
“A lot of products in the past have tried to do intent-based networking,” said Mike Dvorkin, co-founder and CTO of Hedgehog. “Intent is amazing, but what is going on underneath is always the challenge.”
The Hedgehog Open Network Fabric sidesteps a lot of these challenges resulting from technical complexity in the hybrid distributed cloud by making operations hands-off. This makes it a lot simpler for people with no or little networking expertise like the DevOps and SREs to manage and operate the fabric.
“The whole thing is designed to provide a hands-free operation. You bring it into the environment, set up your rack, press the power button on and it just comes up. This way when you’re deploying things at the edge or data edge where you cannot have a dedicated networking personnel, it works,” he explained as he and the team showcased Hedgehog Open Network Fabric to the audience at the recent Networking Field Day event in California.
With Open Network Fabric, Hedgehog has two clear goals in mind: to deliver a network that is nimble and responsive to the customer’s use case – be it AI, ML or data analytics – and to make networking everywhere look and feel like the public cloud.
The Hedgehog Open Network Fabric
The Hedgehog Network Fabric is easy to understand and demands little to no skill. It is a software-driven solution that is deployed on top of Kubernetes. Similar to public cloud, the Fabric is built on VPCs or Virtual Private Clouds. VPCs are private virtual environments hosted within public cloud that provide isolated highly scalable computing environments to customers.
Hedgehog VPC delivers multi-tenancy in hybrid cloud. Tenants can seamlessly run their virtual networks with their unique private address namespaces.
“Those VPCs are like network containers. You build a lot of services around them. We’re peering across VPCs and to the outside and a lot of exciting stuff is coming later,” Dvorkin told.
Operators, through a multi-tenant API, can define network intent for connectivity and isolation, and count on those to get pulled into the configuration of the software appliances and switches.
“It’s super-simple to operate. People who understand networking can dive under the hood and know what’s going on with the network without having to struggle with obstructions,” he said.
Hedgehog does not put customers on a vendor lock-in when it comes to appliances. Any whitebox or graybox Broadcom-based switches, SmartNICs/DPUs and CPUs from any vendors work.
The software on the switch too is not proprietary, but an open-source NOS, SONiC. SONiC comes pre-bundled with Hedgehog, but customers have the additional flexibility to choose other distributions.
The Hedgehog Fabric uses Kubernetes API to control and manage its resources.
“The way we’re treating the network because of the Kubernetes control plane, would basically turn the entire Fabric into a huge Kubernetes cluster. So every switch, every NIC or DPU, every gateway becomes part of the cluster.”
Kubernetes has a reputation of being notoriously complex. To ensure that users do not have to experience the operational complexities, Hedgehog abstracts away all management tasks.
“If you don’t want to struggle with Kubernetes and learn new things, it’s completely hidden away from you – completely self-managed.”
Hedgehog’s networking infrastructure services include zero-touch provisioning. ZTP makes deployment super-simple. Operators can design the fabric based on what they need, and following the diagram, Hedgehog bootstraps and configures the devices automatically.
“Each of the switches that we support, there is a piece of metadata that normalizes it so we can work with it.”
Dvorkin highlighted front-panel booting, a feature that lets switches to be booted on front-panel ports. This way of booting eliminates the need for a separate management network. Chainbooting on Hedgehog allows operators to boot the network from the nodes that are up. Down the line, the company plans to add switch-level caching to allow nodes to be booted from neighbor nodes.
Hedgehog supports multiple topologies – collapsed core, CLOS networks and small mesh environments.
“If you have a small enough network and you don’t want to waste money on spines, but just hook things up in the mesh, you can hook up gateways to that and watch your compute gear,” he said.
Hedgehog does not rely on a separate out-of-band network. Everything is managed in-band.
“We police and control management and protocols traffic because you’re now sharing things with tenant networks. We also do full-on isolation and not just from a performance perspective, but also from security perspective.”
Going forward, Hedgehog will roll out new features and functionality making it a complete networking platform.
Be sure to check out Hedgehog’s presentations from the recent Networking Field Day event on the Tech Field Day website to learn more about the platform.
