IEEE 802.3ad is a great way to remove spanning tree from your network. However, IEEE 802.3ad doesn’t work very well if one end of the bundle is split across two routers. Multi-Chassis Link Aggregation (MC-LAG) is a protocol that allows two routers to appear as single logical router to the other end of the IEEE 802.3ad bundle.
The most typical use case for MC-LAG in a Service Provider network is to provide customers both link-level and node-level redundancy. A good side effect of MC-LAG is that it removes the need for VPLS multi-homing (MH). For example, if a Service Provider had 4,000 VPLS instances that required node-level redundancy, one solution would be to implement VPLS MH; however, if there were a node failure, all 4,000 VPLS instances would have to signaled to move to the redundant PE router. The alternative is to use MC-LAG to provide node-level redundancy and eliminate 4,000 instances of VPLS MH; this method fails over the entire IFD in a single motion instead of every single VPLS MH instance.
Enterprise environments find that MC-LAG is a great method for multiple core routers to provide a single, logical IEEE 802.3ad interface to downstream switches and avoid having spanning tree block interfaces. From the perspective of a downstream switch the IEEE 802.3ad connection to the core is a single logical link, but in reality there are multiple core routers providing node-level redundancy.