Networking

Cisco ACI Data Center Design

A practical look at spine-leaf topology, centralized policy control, and the integration of network, compute, and storage domains.

Cisco ACI fabric topology connecting spine and leaf switches to compute and storage domains
Cisco ACI fabric topology with APIC policy control, compute, storage, and external connectivity.

At the center of this topology lies the Cisco ACI fabric, formed by a spine-leaf architecture. Two Nexus 9332C switches serve as spine switches, and two Nexus 9336-FX2 switches act as leaf switches. These switches are interconnected using high-speed 40GbE links, forming a non-blocking, low-latency fabric.

All data center devices connect only to the leaf layer, never directly to the spine. This creates a scalable, predictable, and programmable network design.

Centralized policy through APIC

Policy control is managed by the ACI APIC cluster. The three-controller cluster pushes policies, manages configurations, monitors performance, and enforces security across the infrastructure. Instead of configuring VLANs and routing manually, APIC provides an application-centric operating model that simplifies day-to-day operations.

External connectivity

External connectivity is provided by a management switch connected to the ACI leaf switches using virtual PortChannel links. This design enables resilient Layer 2 bridging for legacy systems, supporting smoother migrations and secure north-south traffic flows.

Compute and storage integration

The compute domain uses Cisco UCS Fabric Interconnects, connecting northbound to the ACI leaves and southbound to compute nodes. This provides end-to-end automation and policy enforcement while reducing operational complexity.

The storage domain connects directly to the ACI leaves using high-speed port channels. Consistent security and quality-of-service policies can therefore be applied to storage traffic, helping maintain predictable application performance.