i) Cisco ACI Single POD Fabric:
--> All the Leaf nodes are connected in a full mesh manner with Spine nodes in the single pod/data center
--> Requires one APIC cluster running of three nodes to manage the fabric
--> A Single instance of ACI control plane protocol runs between all the Leaf switches and Spine switches in the fabric
--> If you want to build DR/Active DC solution then you need a separate APIC cluster and separate pod in another site
--> Not a scalable solution
ii) Cisco ACI Stretched Fabric:
--> All the Leaf nodes are connected in a full mesh manner with Spine nodes in the single pod/data center
--> There are some leafs in each data center/pod connects to the spines present in other data center/pod
--> These leafs are knowns as transit leaf nodes as the traffic to other DC/pod goes through these leafs
--> Requires one APIC cluster running of three nodes to manage the fabric ( You can have two APIC nodes running in one DC and one APIC node running in another)
--> A Single instance of ACI control plane protocol runs between all the Leaf switches and Spine switches in the fabric
--> Not a scalable solution and creates a single failure domain
iii) Cisco ACI Multi Pod:
--> To prevent single failure domain and provide resiliency in the Fabric, Cisco came up with the Multi-Pod Solution
--> All the Leaf nodes are connected in a full mesh manner with Spine nodes in the single pod/data center locally to the site
--> The Pods/Data Center Fabrics are connected using an external device known as Inter Pod Network Devices
--> Requires one APIC cluster running of three nodes to manage the fabric ( You can have two APIC nodes running in one DC and one APIC node running in another)
--> A Single instance of ACI control plane protocol runs between all the Leaf switches and Spine switches within the pod
--> The main advantage of the Cisco ACI Multi-Pod design is hence operational simplicity, with separate pods managed as if they were logically a single entity
--> Single change domain in ACI Multi Pod ( Whatever the logical configuration such as Tenants/VRF etc pushed to all Pods)
--> Maximum latency of 50 msec Round trip time is required between pods for ACI Multi POD to work properly
--> Active/Active and Active/Standby of Firewall and F5 cluster is supported in this design
iv) Cisco ACI Multi Site
--> The same APIC Cluster is used to manage the ACI Multi Pod setup, Only Control plane isolation provided by Multi Pod
--> To provide complete isolation of the control plane and management plane perspective, Cisco came up from ACI Multi-Site
--> Cisco ACI multi-site provides complete isolation of the control plane and management plane
--> Cisco ACI Multi-Site is also called a Multi Fabric design as it interconnects multiple fabrics using Inter-Site Network (ISN) Devices
--> The Pods/Data Center Fabrics are connected using an external device known as Inter-Site Network Device
--> A Single instance of ACI control plane protocol runs between all the Leaf switches and Spine switches in each fabric/pod
--> All the Leaf nodes are connected in a full mesh manner with Spine nodes in the single pod/data center locally within each site
--> Requires one APIC cluster running of three nodes to manage the fabric in each site
--> Nexus Dashboard Orchestrator is required to manage and implement ISN Configuration required for Cisco Multi-Site
--> Nexus Dashboard Orchestrator is required if we want to integrate Cisco ACI Fabric with Azure/AWS Public Clouds
--> Cisco ACI Multi-Site provides layer2 extension across the sites using VXLAN.
--> Active/Standby of Firewall and F5 cluster is supported in this design
--> We can use Nexus Dashboard Orchestrator to push configuration changes local to the site or across the sites.
Best Regards,
Kareem
Reference: Cisco ACI Muti Site Architecture White Paper