I wanted to share some highlights of vPC also known as Virtual Port Channel. First, let’s remember that vPC is one of a few technologies that fall under the umbrella of MCEC which stands for “Multi-Chassis Ether Channel.” Stackwise and VSS (Virtual Switching System) fall into the same category but don’t work on the same equipment nor work in exactly the same way.
I’d like to also point out that the vPC headend or peers are usually done on Cisco Nexus equipment.
With vPC, we can make two Nexus devices peers. They then appear as one Layer 2 switch to the downstream device. The result is that a downstream device like a switch or a server, can have a two (or more) port port-channel which happens to spread over both of those upstream peers. The downstream device sees a single port-channel upstream and vPC makes Layer 2 think that those are both one switch.
Thus, the two (or more) uplinks are not seen as separate paths for the purposes of spanning tree, so neither one is blocked. This way, the downstream device can use ALL of the bandwidth to the two upstream devices and still have automatic failover if one of the paths has a problem because port-channel (using LACP) does that by default.
In order to allow both upstream Nexus devices to have their paths open to the downstream switch, they communicate and cooperate over what is known as the “peer link” which must be at least 10 Gigabits Per Second.
vPC can keep from blocking one of the paths and STILL not have a loop because it has a built-in mechanism stopping traffic which came up from a particular vPC number on one machine from going back down on the same vPC number in the other machine.
Two Route/Switch entities must create a peer relationship but not more than two. A peer can only belong to one domain at a time and can only have a relationship with one peer.
The term “Route/Switch entity” is used because Nexus 7XXX series switches can have VDCs (Virtual Device Contexts) and each VDC is independent of the others which allows each of them to have their own vPC peer/domain relationship. Two VDCs on the same N7K may not have a vPC peer relationship with each other.
Finally, one of the coolest things about vPC is what it does when the vPC peers are also acting as the Default Gateway for a subnet. With vPC, a protocol like HSRP or VRRP can share the VMAC of the redundancy group simultaneously, thus allowing both peers to process outbound traffic instead of having to switch everything to one of the two who is the VRRP or HSRP active.
So that’s a quick “lowdown” on vPC. Are there more details? Sure. But I hope this serves as a quick highlighting the highlights!