Selector targets the network visibility gap in multi-cloud infrastructure

Enterprises aren’t just moving applications to the cloud. They’re moving the network itself.

An increasing number of organizations are shutting down data centers and migrating networking infrastructure directly into the cloud. BGP sessions, virtual firewalls, transit gateways, and VPN terminations are following the compute. The challenge for many organizations is having proper observability across both on-premises and multi-cloud networks.

That’s the challenge that NetOps vendor Selector is now tackling with an update this week to its platform. The new capabilities give network teams a single correlated view across branches, colocation facilities, on-premises data centers, and public cloud infrastructure.

“I need the same type of detail, AIops and observability for my network constructs in the cloud that I had to do for my data center, connecting to my branch locations,” Kannan Kothandaraman, co-founder and CEO of Selector, told Network World.

Closing the visibility gap

Selector is positioning the new multi-cloud capabilities as an extension of its existing platform, not a separate product. The goal is to let network teams trace a connectivity problem across the full hybrid path without switching tools or context.

That path can span an SD-WAN provider, a service provider circuit, a colocation interconnect, a Direct Connect gateway, and a VPC transit gateway before reaching a cloud-hosted application. Kothandaraman said most APM tools do not go deep enough on network path analysis to cover that end-to-end trace.

Early deployments have surfaced issues that were not previously visible. In one migration review, Kothandaraman said that the platform revealed an application flow the customer did not know existed.

“If my branches are complaining that I cannot reach an application, is it a problem with my SD-WAN? My service provider circuit? Am I having a problem with my Equinix, you know, direct connect gateways? Am I having a problem with my VPC transit gateways? “Kothandaraman said. “Somebody needs to stitch all of that together. That’s the solution we put together here.”

How the platform works

The foundation of Selector’s approach is a normalization layer the company calls the data hypervisor. It sits between the telemetry ingestion layer and the AI and ML engines above it. Its function is to make incoming data source-agnostic before it reaches the analytics layer, regardless of where it originates.

In cloud environments, ingestion is primarily API-based. The platform collects VPC flow logs and subscribes to hyperscaler event streams for infrastructure change data. It also pulls telemetry from third-party tools including virtual firewalls and load balancers.

On the on-premises side, SNMP and streaming telemetry remain in the mix. Cloud-native constructs covered include AWS transit gateways, Direct Connect gateways, and Google Cloud Routers. The platform also handles virtual VRFs and BGP configurations within cloud environments.

One capability specific to cloud is tracking changes on the hyperscaler side. Enterprises do not control the underlying physical infrastructure. When a hyperscaler publishes a change event, the platform correlates it with observed network behavior to determine whether it is causing an issue on the customer side.

Kothandaraman noted that cloud telemetry proved more straightforward to normalize than on-premises data. On-premises environments carry significant vendor variation in data formats. In cloud environments, hyperscalers publish structured event data the platform can consume consistently.

What’s next

The data hypervisor that normalizes telemetry across on-premises and cloud domains is also the foundation on which Selector is building its next layer. The company has been developing foundational AI models for network infrastructure since it was founded, with generative AI incorporated more recently across the product. The initial focus has been on the user experience layer.

Kothandaraman said the company has taken a deliberate approach, building foundational infrastructure models first and layering generative AI on top of them. A full generative AI release covering the entire platform is planned for the fall.

“We have a major launch coming up in the fall, where we will reveal our full, generative AI-based transformation and solution on the product side. It’s eye-popping, what you can do,” he said.