A Network Switch’s main purpose is to provide connectivity between LAN clients. Any switch that provides this ‘basic’ functionality, operates at the 2nd OSI layer also known as the Datalink Layer.
As we look at more intelligent operations, such as InterVLAN routing, routing protocol support and more, we start moving upwards the OSI Model.
More specifically a layer 3 switch provides functions based on the 3rd layer of the OSI model, this includes routing between networks, VLANs and other similar services. Today’s complex and highly intelligent switches can operate up to layer 7 of the OSI model, performing IP routing services, Quality of Service ( QoS), firewall functions, deep packet inspection and much more.
Example of such switches are the Cisco Catalyst 6500 series that can be populated with various modules to increase their functionality and supported services. These type of switches are usually found in Internet Service Providers and large datacenters and are made to handle huge amounts of traffic.
As a rule of thumb, the higher OSI layer a switch is said to operate at, the more intelligent is it. This also means that the switch has more powerful CPU(s) , greater amounts of RAM and in most cases, requires more power to operate!
When designing the network infrastructure, we tend to use ‘smarter’ switches at the network backbone (Core switch) and less smarter switches at the ends – where users connect to.
Hope that helps make things a bit clearer!
Previous Member’s Answer:
A switch can operate at both layer 2 (data link) and layer 3 (network). Layer 2 uses MAC addresses for switching. Layer 3 uses IP addresses for switching. Take a look at this wikipedia article on network switches for additional info.