Cisco ACI 2.0: Multi-Pod Fabric

Today I am going to talk about Cisco ACI 2.0 Multipod Fabric infrastructure and the components involved. Cisco ACI is one of the demanding technology in the market for data center policy based environment infrastructure.

The basic evolution of ACI stretched fabric is multi-pod. The control protocols that run inside the fabric (ISIS, COOP, MP-BGP) are different instances at different sites, which means failure domains are now split between pod’s contributing to increase the overall design resilience.

Fig 1.1- basic design of Cisco ACI Multipod
Even though control domain are separated we still have same change-domain, meaning configuration or policy definition applied to any of the APIC nodes would be propagated to all the Pods managed by the single APIC cluster as shown in Figure 3 below.
The different Pods may represent different islands (rooms, halls) deployed in the same physical data center location, or could map to geographically dispersed data centers (up to 10 msec RTT latency).
Different workloads that are part of the same functional group (EPG), like Web servers, can be connected to (or moved across) different Pods without having to worry about provisioning configuration or policy in the new location.
At the same time, seamless Layer 2 and Layer 3 connectivity services can be provided between endpoints independently from the physical location where they are connected
From a physical perspective, the different Pods are interconnected by leveraging an “Inter-Pod Network” (IPN). Each Pod connects to the IPN through the spine nodes; the IPN can be as simple as a single Layer 3 device, or can be built with a larger Layer 3 network infrastructure.
  • IPN devices allows for the establishment across Pods of spine-to-spine and leaf-to-leaf VXLAN tunnels. There are few basic requirements for a switch platform to be chosen as IPN
    Platform should support PIM-BIDIR. This helps BUM traffic to be exchanged between endpoints in different data-centers and also allows to build a source specific multicast tree within fabric.
  • Should support handling jumbo frames. Jumbo MTU support (9150B) to handle VXLAN encapsulated traffic
  • Support for DHCP relay
It is the use of the VXLAN overlay technology in the data-plane that provides seamless Layer 2 and Layer 3 connectivity services between endpoints, independently from the physical location (Pod) where they are connected.

Running a separate instance of the COOP protocol inside each Pod implies that information about local endpoints (MAC, IPv4/IPv6 addresses and their location) is only stored in the COOP database of the local spine nodes.

Since we need to provide consistent view of endpoints across all sites an overlay control plane is running between spines and used to exchange reachability information. The overlay protocol used to exchange L2 and L3 info is MP-BGP.