Bridges

2.2.2 Bridges
While hubs are used to build a network at a single site, bridges are used to build a network at two sites—or to join two networks . A bridge operates by looking at the header of the data that comes to it. If the data is for the network on which the bridge resides, the bridge leaves the data alone. If the data is for another network, the bridge gets rid of the data by sending it to a predefined location. An example would be two networks, one in New York and the other in Chicago, that have a bridge at each location (on each network). If a user sends a message in New York and it is not for another user in New York, it must be for a user in Chicago, so it is sent there. Likewise, if a user in Chicago sends a message and it is not for another user in Chicago, it must be for a user in New York.

A bridge can never be used with more than two sites. If San Francisco were added into the mix, the bridge at Chicago could not determine whether to send the message to San Francisco or to Chicago and could send it to only one location.
The biggest advantages to bridges are that they are reasonably cheap, and they work with all protocols by dropping down to and concentrating on the physical addresses of devices. Physical addresses give bridges the ability to work with NetBEUI (NetBIOS Extended User Interface) and non-routable protocols as easily as they work with TCP/IP. Remote bridges are nothing more than bridges that connect two LANs into a WAN and filter signals.
It is important to understand that a bridge—like a hub—receives every data packet sent on the network. The bridge then looks at the header and at an internal table (known as the forwarding database, or routing table) and determines if it should leave the packet alone or send the packet out to the address it has. In this capacity, a bridge is used to expand the geographic scope of the network to another location. The opposite could also be true in that a bridge could be used to divide one network into two segments to reduce traffic throughout the whole network.
To visualize the latter situation, suppose that a company has two large departments: Manufacturing and Sales. Every piece of data generated by Manufacturing is sent throughout the network, as is every piece of data generated by Sales. If the network could be divided into two segments with a bridge between Sales and Manufacturing, the network traffic could arguably be cut in half. All the Manufacturing traffic would stay on the Manufacturing segment, and all the Sales traffic would stay on the Sales segment. Data would cross over through the bridge only when Sales requested data from or sent data to Manufacturing, or vice versa.