Comparing Layer 2 with Layer 3
When you first enter the networking world, Layer 3 switching can seem too complicated for its own good. Introducing segmentation across the network and increasing overall complexity, with additional protocols required to handle the distribution of routing information just so two or more switches can exchange traffic… this all makes Layer 2 networking child’s play by comparison!
Yet this additional complexity brings more flexibility. It enables loops and multiple paths through the network to be fully utilised, thereby increasing throughput and redundancy while decreasing latency. There’s also the possibility of using protocols such as OSPF to further optimise traffic flow through the network.
A common example of Layer 3 network architecture’s superiority to Layer 2 network architecture is the implementation of a Clos network. This is a type of Full-Mesh network consisting of two layers of switches, a high throughput core (called Spine switches) and Top of Rack distribution switches (called Leaf switches). The structure of a Clos architecture means both link saturation and redundancy are possible. The nearest L2 facsimile of this would be a tree, where in order to enjoy any form of redundancy a spanning tree protocol would have to be deployed forming a very complex and strict hierarchy tree over your network, where any single interconnect can become a severe bottleneck affecting the whole network.