Fast forward three months to September 2024, and Cisco has reached beta/early availability for Nexus Hyperfabric, although not yet for the AI-centric flavor of Nexus Hyperfabric. Details now reveal the broader strategy for Nexus Hyperfabric, and it goes far beyond what was described during the opening act. While the AI-centric version continues to develop, Cisco intends to deliver Nexus Hyperfabric for various use cases, including re-platforming existing modest-sized data centers for easier cloud-managed operational lifecycles, edge/remote data centers (such as light-touch smaller fabrics in hospitals, manufacturing and power sites, retail offices, development sites, and transportation hubs), and fabrics deployed in colocation facilities. These scenarios have not been ideally suited for the existing Cisco Nexus switch product lines, which are designed for on-premises operational models and rich customization options. In order to deliver Hyperfabric, the Cisco team has developed a new switch and fabric architecture that has very new and different characteristics:

Cloud-managed. Cisco’s new Cloud SaaS Controller facilitates design, including automatically prescribing a full bill of materials (equipment, cabling, power, etc.) and a deployment plan, automates configuration (switches “phone home” when connected and are automatically configured by the controller), automates validation testing to ensure consistent application of configurations and policies, and then continuously monitors the switches for health, activity, and performance. Cisco has built the UI using the same framework as the Meraki Dashboard, a proven and highly rated cloud network management solution.

Open and remotely programmable. The new 6000 series switches are Cisco’s first to support SONiC standard networking, supporting open network architectures. Although built on SONiC—Cisco is a consortium member—customers will not have direct OS access on these switches. The Cisco 6000 switches are only configured and managed utilizing the Cloud SaaS Controller, using either the web-based console or a full JSON API, the latter enabling integration with DevOps tools such as Ansible and Terraform. This means that there will be no command-line interface, and no reason—or means—for opening direct sessions on each individual switch.

Easier to use. The new Nexus Hyperfabric is a switching solution intended to meet the common needs of the many rather than the unique needs of the few. The fabric underlay is an EVPN and VXLAN fabric which fully supports dual-stack IPv4 and IPv6, but these details are largely abstracted away from the user. Configuration is limited to a simple Layer 2 or Layer 3 setup, and all configuration details are determined and deployed by the Controller. The design objective is to enable a broader range of IT pros to engage in building, deploying, and operating the solution, and not only those with deep networking expertise.

Scales gracefully. Target deployments can utilize leaf-spine architectures as well as simple mesh configurations, providing flexibility that is especially vital for distributed or edge data center deployments, such as transport hubs, hospitals, and power and manufacturing sites. These deployments can scale down to a single standalone switch but importantly, if more capacity is needed, additional switches can be added and will be automatically recognized and incorporated. This means that it will no longer be necessary to pre-plan and deploy a complete network fabric to anticipate future computing needs—rather, networking can be added as needed along the way, just in time.

Integrated services. Cisco owns and manages the cloud controller and ensures that the underlay fabrics are properly configured through automation. Customers are responsible for the devices that connect to the fabric, such as hosts and upstream routing, and the networks they use. There are no on-premises or air gap options at this time, although those are under consideration for the future.

The Nexus Hyperfabric is targeted at the growing number of distributed smaller data and compute clusters that are commonplace in geographically diverse organizations, including increasing interest in edge computing stacks. It is not meant to be an alternative to networking solutions for large core data centers, where Nexus switch fabrics are still the primary offering and continue to support all of the customizations necessary for purpose-built, highly tuned and optimized connectivity. That said, it is possible to connect Nexus Hyperfabric to other Nexus fabrics, by installing a Hyperfabric switch in the non-Hyperfabric data center as a gateway.