Pluribus Networks: Disaggregated Networking for Highly Distributed Cloud Environments
Data Centre ICT Automation/Orchestration Innovation of the Year
As 5G, AI/ML and edge compute drive data center (DC) locations to become more distributed, miniaturized and often “lights-out”, virtualizing and automating DC networks to increase agility and automate operations becomes an imperative. Networking has always been the DC automation laggard and an impediment to delivering the agility needed by DC operators in the modern cloud era. This has been the case in large centralized and regional DCs and is only exacerbated with highly distributed and space/power contained DCs located in central offices, at base stations and in enterprise edge locations.
The team at Pluribus Networks is delivering a new and highly innovative approach to this emerging networking challenge, leveraging the principles of disagreggation, network virtualization and controllerless SDN to deliver a data center network fabric optimized for highly distributed cloud environments.
The Netvisor ONE operating system (OS) is designed to run on bare metal switches delivering the cost, scale and speed-of-innovation benefits delivered by disaggregation. Netvisor ONE runs on white box switches from multiple hardware providers and is the exclusive OS to support Celestica’s Edgestone, the industry’s only NEBs-compliant DC switch designed for central office environments.
In turn, Netvisor ONE powers the Adaptive Cloud Fabric (ACF), a controllerless SDN solution that provides a fully automated underlay and overlay fabric that virtualizes the DC network to provide cloud-like agility, multi-tenancy, network slicing, per-flow telemetry and multiple orchestration integrations. Federating up to 1,024 switches across any set of geographically distributed DCs, ACF enables this fabric of switches to act like a single logical Layer 2 / 3 switch with a single management IP for radical automation. DC operators can easily move workloads from one DC to another (VMs or containers) and share DC resources for maximum efficiency. The underlay and the overlay network are automatically deployed via the powerful UNUM Fabric Manager GUI and a single command will consistently instantiate a network service across the entire fabric with ease. In turn multiple fabrics can be federated together with EVPN to distribute network services on a per tenant or per slice basis across multiple fabrics.
The SDN automation requires no external controllers - lowering licensing and hardware cost and reducing space/power consumption. Pluribus has taken a completely different approach to achieve this, developing innovative and compact software that leverages the embedded compute and memory footprint contained within the DC switches themselves to create a distributed, controllerless SDN implementation. Think of each switch acting like a server and containing a mini-SDN controller and a distributed SDN database for the underlay and overlay network.
The value is clear and the solution has gained tremendous traction. Pluribus Networks is embedded in the Network Function Virtualization Infrastructure (NFVi) of a major NEM and now has more than 100 mobile network operators that have deployed Netvisor ONE and ACF in mission critical 4G/5G virtualized mobile cores around the globe. Pluribus has more than 350 customers worldwide and is growing rapidly as the value of this approach is understood by an increasingly broader set of customers.
For example, imagine there is an 5G operating area with 20 edge data center sites with each site containing 6 switches - these 120 switches would be formed into a single, automated and sliceable multi-tenant fabric.