Picture this in your mind: long, freshly paved blacktop highways going for miles. In a group of adjacent street blocks, the pavement makes the intersections into one seemless city. You drive along, blissfully unaware that you have passed from one city to the next as you travel down a major city street. In the bigger scene, you take Interstate 5 to San Diego, leaving the collective cities of Los Angeles and Orange counties behind, but still seamlessly approaching another group of collective cities in San Diego county.

Have you ever wondered about how your computer's network can talk to another network? Internetworking is the latest and greatest of manifestations of networking technology, and the connections are very much like the highway example. Bridges are like the streets in a city. They connect separate networks into a single logical network, much the same way as a major street like Pacific Coast Highway traverses several cities. A router, however, is like a freeway where each network retains its logical identity as a separate network segment, much the same way as Los Angeles county and San Diego county are separate, but still connected.

The technical description of the differences between bridges and routers is defined in terms of the Interational Standards Organization (ISO) Open Systems Interconnection (OSI) model. The OSI model describes an internetwork as consisting of seven (7) layers, from logical to physical. This would be like have level 7 being Earth (the logical layer), level 6 being the United States, level 5 being California down to level 1 being our streets and highways (the physical layer). This interationally recognized model for data communications serves as a basis for understanding and characterizing an overall internetworking strategy. This is the relationship of bridge and router services to the OSI model:

A repeater, which operates only at layer 1, the physical layer, regenerates signals and permits extended network length. It is analagous to getting another tank of gasoline to make the trip to San Francisco. A gateway is a protocol conversion device. Operating at any layer of the OSI model, it translates between two different peer protocols within a network. (Protocols will be explained later.)

Bridges do most of their work at layer 2, the data link layer, while routers work at layer 3, the network layer. This difference is important. The layer at which each type of system functions affects both its capabilities and operation. Bridges "see" the network in terms of device addresses only. They use device addresses as the basis for the decisions they make about handling packets, the basic units of information that are transmitted throuh a network. Information about paths, or routes, through the network is not accessible to bridges because such information is encoded in the network address, which is only accessible to a system operating at the network layer. Thus, bridges do not make decisions about paths through the network. As a result of this limited decision making capacity, bridges are relatively simple devices. They can provide an attractive and inexpensive way to internetwork.

Routers, on the other hand, "see" a network both in terms of network addresses and paths. Routers "know" all the paths between any two points on the network, and they know which of these paths is the shortest. They may also know other characteristics of each pathway, such as its operational status, its bandwidth, or its economic cost.

Because of the additional information available to them, routers inherently can do more things with packets than bridges. As a result, routing software is more complicated than bridging software, and is therefore more difficult to develop and implement. Today, bridges and routers are often based on the same hardware platform. Bridging and/or routing functionality is provided by the software with which these hardware systems are configured. This approach offers a great deal of flexibility to network managers. It means that the system used to create each network interconnection can be easily adapted to fit changing circumstances.

If bridges and routers are the politicians in internetworking, then protocols are the laws devised to establish and maintain communications between points on the network. Rules that are standard across different types of equipment manufactured by different vendors are standard protocols. Like bridges and routers, protocols correspond to layers of the OSI model. To perform the full range of network functionality, several protocols will usually be active simultaneously in any given network. (i.e. People will be on the city streets and the freeway at the same time.)

In most cases, the multiple protocols that exist in an internetworking environment are related to one another. They are members of what is known as a protocol suite, sometimes referred to as a protocol stack. There are many protocols and protocol suites. Some of them are proprietary and were originally designed to run only on equipment produced by the vendors who defined them. Examples of proprietary protocols and corresponding vendors include, for example: DECnet from DEC, IPX from Novell, SNA from IBM, and XNS from Xerox. In today's internetworking world, these popular proprietary systems have become the standards supported by many vendors. Moreover, they normally include system interfaces to nonproprietary, vendor independent standard suites.

Two protocol suites were developed from the start to be vendor independent standards: the Transmission Control Protocol / Internet Protocol (TCP/IP) suite and the Open Systems Interconnection (OSI) suite. The TCP/IP suite has attracted a large following in the internetworking marketplace. TCP/IP standards are specified by the Internet Activities Board (IAB), a broad consortium of corporate, academic, and governmental organizations. Like the model of the same name, the OSI suite is defined and approved by the ISO. Accordingly, when people refer to either the TCP/IP or the OSI protocol stack, they are referring to standards based internetworking systems.

There are two basic types of bridges and routers, local and wide area. The difference between them depends on their network interfaces, or ports. Local devices are equipped with ports that permit connection to network backbones. Typical of such media are coaxial, fiber optic, and twisted pair cable. An important attribute of local devices is their ability to connect networks that use different media. For example, they can connect a coaxial network with a fiber optic network, or they can connect either with a twisted pair network. Wide area devices, on the other hand, are those with communications interfaces compatible with long distance transmission media. They usually have two or more long distance communications ports and at least one local port, and can connect networks situated across town or around the world.

There are two basic types of long distance technology: point-to-point links and "cloud" technologies. Point-to-point links are typically lines leased from telephone or other communications companies; these lines can range from low to very high line speeds. Common examples of point-to-point lines speeds are 19.2 Kilobits/second (Kbps), 64 Kbps, and T1/E1 (1.544/2.048 Megabits/second).

Cloud technologies are switching systems that route information through networks in a way that is totally transparent to users of networks attached to the cloud. To those users, the operation of the cloud looks like a point-to-point connection even though the information may actually have traveled over several different communications lines. The most common interfaces to clouds are X.25 and frame relay.

Now lets take a look at how bridges work. Bridges provide a way to join two or more networks together to form a single logical network, and they accomplish this in a way that is transparent to every device on the network. The original networks are referred to as network segments in the resulting network. Both local and wide area bridges regenerate the packets that they forward. Neither the number of bridges through which a packet travels nor the distance it travels has any effect on signal quality. So one significant use of bridges is to extend the distance an internetwork covers.

Bridges deal with both source and destination device addresses. The source address is the address of the device that initiates a transmission of information. The destination address is the address of the device to which the information is bound. This distinction between source and destination device addresses is of central importance to the operation of bridges. Briefly stated, bridges "learn" on the basis of source addresses and "forward" on the basis of destination addresses. Learning and forwarding, along with a third process called "filtering" and an additional refinement known as the Spanning Tree Algorithm, constitute the basic functionality of all bridges.

Bridges use source and destination addresses to build a database of device addresses. All basic bridge functionality involves transactions with this database. When a bridge receives a packet, it compares the source address with the entries in its address database. If the source address is not already in the database, it is added by the bridge. In this manner, the bridge learns the addresses of the devices on the network. Because of this learning capability, new devices can be added to a network without having to reconfigure the bridge. The bridge then compares the destination address with those in its database. If the destination address is on the same network segment as the source, the bridge "Filters" the packet. In other words, it automatically discards it. This process helps prevent the overall network from getting congested with unnecessary traffic.

If the destination address is in the address database and not on the same network segment as the source, the bridge determines which of its ports is associated with the address and forwards the packet to the appropriate port. If the destination address is not in the address database, the bridge forwards the packet to all its ports except the one on which it was received. This process is known as "flooding." Flooding guarantees that a packet with an unreconized destination will reach all network segments and hence the destination address.

This flooding, however, can be a severe problem for bridge based networks, one that can lead to unnecessary and indefinite duplication of packets. If there are many active loops, or different paths to different network segments, then the resulting excess traffic can degrade overall network performance and cause some protocols to simply stop working. This problem has been recognized and is addressed with an approach known as the Spanning Tree Algorithm (STA). A spanning tree is any unique device-to-device path in the network. The STA constructs a spanning tree through a series of bridge-to-bridge negotiations to determine the path that will remain enabled for transmissions and the path or paths that will remain temporarily disabled. As a result, selected ports on a bridge are placed in a "forwarding" state and the remaining ports are placed in a "blocking" mode. This process guarantees a single path between any two devices on the network.

If the unique path fails for any reason, the bridges participating in the STA activate appropriate blocked ports to create a new spanning tree. Thus, the algorithm allows internetworks connected by bridges to have some path redundancy without the problems associated with active loops. The network can recover quickly and automatically if a bridge or a section of network cabling fails.

Bridges can play a substantial role in managing the resources and facilities of complex networks. Insofar as they receive all traffic on each attached seement, bridges are "privileged observers." They are ideal sites at which to collect general network statistics. The statistics might include, for example, percent network utilization over time, number of packets transmitted, and the number of collisions or alignment errors. This information is normally handled in one of two ways. It can be immediately displayed on a locally connected terminal. Alternatively, the bridge can function as a network management "agent." A network management agent is a system running the "agent" software of a network management protocol such as the Simple Network Management Protocol (SNMP), which is part of the TCP/IP protocol suite. An SNMP agent forwards the collected information to a central SNMP network management site known as the network management station. These statistics could prove vital to network managers seeking to optimize network performance.