![[previous]](../../../graphics/l_arrow.gif){kind=icon}
![[next]](../../../graphics/r_arrow.gif){kind=icon}
During the initial configuration of the OfficeConnect Remote 840, you must decide whether to configure the unit as a bridge or as a router. If you are unsure which option you should choose, this section will help you decide.
Bridges and routers are used to connect networks together. The cost of connecting networks together is generally proportional to the distance over which the network extends and the amount of bandwidth required. Large amounts of bandwidth can be provided easily within a LAN by connecting different segments together with a local bridge. However, it becomes impractical and expensive to extend this bandwidth over larger distances, and it is, therefore, usual to interconnect local high-speed networks using bridges or routers connecting over slower speed terrestrial and satellite links.
In the following sections we describe the concepts behind bridging and routing, and discuss the different ways in which LANs can be configured and operated to optimize performance and minimize disruption of traffic on each individual LAN.
A bridge connects one or more LANs together. It examines each data frame received at a LAN port and forwards any frames that it assumes are for a destination device not connected to that LAN port. The bridge is able to do this by learning which devices are connected to each LAN port.
A router learns much more about the networks connected to it and is able to be much more selective about the data it passes on to other networks and to which network it transmits. By default routers reject or filter data unless it matches predefined attributes (for example, specific protocols or destination network addresses). In large interconnected networks, a router selects the best route for data to travel.
The list below outlines some of the reasons you might choose to configure the OfficeConnect Remote 840 as a bridge or a router. Read through the rest of this section for more explanation and to help decide which of the above conditions apply to your network.
When a bridge is first powered on, it does not know the number or the locations of stations that are connected to the LAN. To minimize the amount of data passed over the bridge, it must learn the whereabouts (address) of stations to ensure that it passes only the data that is necessary to be passed over the bridge.
Like the envelope of a letter, the header of each frame of data transmitted on the network has a From (source) address and a To (destination) address. This ensures that data reaches its destination on the LAN and that the receiving station can reply. The bridge reads every frame of data received at the LAN port and extracts the source address of the frame. From this information it builds an address table of stations it knows to be on the LAN.
To decide if data should be passed over the bridge, the bridge examines the destination address of the frame. If the address is already in its address table, the bridge knows the destination is on the LAN and therefore discards or filters the frame.
If the destination address is not in the address table, the bridge transmits the data across the bridge. It does this even if the destination device is on the local LAN because it does not recognize the destination station as local. However, if the destination device is on the local LAN, once it replies to the original source station, its own source address is part of the data frame and it is learned by the bridge and added to the address table.
By operating in this way, the amount of data forwarded by the bridge is kept to a minimum. Traffic that is for devices on the attached LAN is rarely forwarded over the bridge.
A bridge can be configured to forget or age a station's address after a period of inactivity, a facility that is used to ensure that stations that are no longer attached to the LAN, do not remain in the bridge's address table, using up space that may be required for other station's addresses.
Some bridges allow address information to be manually configured into the bridge, provided the automatic learning facility is turned off. This will not normally prove necessary unless specific traffic filtering is required.
You can also configure a number of other features to improve the performance and operation of the OfficeConnect Remote 840. These include sophisticated filtering techniques so that only certain types of frames, or those associated with particular work groups, are passed between specific segments.
The OfficeConnect Remote 840 is able to send frames between LANs that may be separated by considerable physical distances. It achieves this by making use WAN links. WANs can be established by using either digital leased lines, ISDN lines, or analog (modem) lines and are usually operated by telephone companies or other service providers.
Figure A-1 shows two LAN segments, A and B, which are connected by a pair of OfficeConnect Remote 840 units, 1 and 2. The type of link between the two depends on the WAN services available at each of the remote bridge locations, and the price the network administrator is willing to pay for those services.
Figure A¯1 Simple Remote Bridging
The OfficeConnect Remote 840 uses RFC 1483 or PPP encapsulation to connect with other OfficeConnect Remote 840 or third party devices.
Large networks of interconnected LANs can be established by using multiple bridges as illustrated in Figure A-2.
The bridges build up their address tables. In Figure A-2, Bridge 1 examines packets from its WAN ports. If the destination unit is not registered as being accessed via the bridge's LAN interface, the frame will not be placed on LAN A. Therefore, frames passing between LAN B and LANs C or D will not impact the overall performance of the LAN.
Figure A¯1 Multiple Remote Bridge
With only a single physical path between LANs, the network is susceptible to link and bridge failures. In the event of a failure, the connection between any of the LANs upstream or downstream from the point of failure will be broken. A more resilient network of interconnected LANs can be established by providing more than one link between any two of the LANs.
Normally, this network would soon encounter serious problems resulting from a loop, around which frames could endlessly travel if precautions aren't taken by the bridges. To prevent loops, you can enable the Spanning Tree Protocol (STP).
When STP is enabled, the bridges send out frames to inquire if there are other bridges on the network. By exchanging information, the bridges block ports that cause the loops and ensure that there is only ever one active path through the network. If one of the links or bridges fail, the other bridges detect this and reconfigure their ports so that there is once again an active data path through the network.
If your network topology is star shaped, a combination of analog modems and bridging is usually the most efficient and successful option. Routing is a better solution if your network is a complex mix of WAN interconnects and/or multiple protocols.
Bridges are programmed to forward data packets automatically by default while routers filter data packets by default. These attributes have an impact on the overall flow of data across the network. Much has been made of broadcast storms in connection with bridged networks, where the broadcast signals from bridges propagate to fill all of the wide area bandwidth, and bring the network down. Broadcast storms cannot be attributed to installation of bridges or routers, but by poor protocol implementation and network design. However the deployment of routers can effectively firewall one logical network from another.
Bridged networks use Spanning Tree Protocol (STP) to provide network resilience, by retaining redundant links on standby, in case the primary link fails. This means that you are not making maximum use of available resources.
Routing protocols make each node aware of the primary and alternate routes available, ensuring that resources (particularly WAN links) are not wasted.
Routers have been designed to provide the optimum route through the network from the workstation through to the destination resource with which the user wishes to communicate. In a very large network there could be multiple paths available, and these could change as links go in or out of service. These changes in network topology are handled by routing protocols.
Some organizations are structured into departments determined by the physical layout of their work environment, so it is natural to divide the corporate network into separate logical networks. Routing becomes the obvious candidate for handling these individual LANs.
The protocol adopted by the Defense Data Network (DDN) for the Internet, is based on obtaining and abiding by, a registered Internet address range. This makes a router the ideal choice for accessing the Internet. Unfortunately, new applicants are only likely to get a Class C registered Internet address, preventing more than 254 connections on one bridged IP LAN.
Running a bridged network allows workstations to communicate directly between one another. A PC user wishing to communicate with a remote network server is totally unaware of any intervening bridges. This is known as transparent operation.
Figure A¯2 Example Network
It is important to understand that in a bridged network the addressing structure for IP relates to a single network. If the units above were bridges and not routers, then an IP node on LAN A could, for example, have an address 140.56.10.1, the node on LAN B an address of 140.56.10.2, and the node on LAN C, an address of 140.56.10.3. All the nodes, therefore, are able to share the same Class B network address, regardless of their location on the bridged network.
Figure A¯3 Open Systems Interconnection Network Layer Model
A routing environment allows stations to communicate indirectly. Following the example in under "Routing IP", let us assume that a station on LAN 1 wants to communicate with a network server on LAN 2. The station on LAN 1, constructs a Layer 2 datalink header (see Figure directly above), with the source station's hardware address, and also the destination hardware address of the local router. To direct the packet to its final network destination, the source station must complete the Layer 3 network header with the destination network address of LAN 2.
Once the packet is received by the Router A, attached to LAN 1, it strips off the network header (refer to Figure above) and examines the Layer 3 header information. It then reviews its routing tables in order to establish where to forward the data packet. It is possible that the LAN 1 router has multiple outgoing ports that would allow different transmission routes to the destination network. In our example using Figure A-3, a packet could go via Router D to get to Router B, or it could go more directly across a single direct link between Router A and Router B.
Figure A¯4 Data Packet Containing Hardware and Software Addresses
The local router contains, within its routing table, information that will allow it to determine the best data transmission route. The type of information the router uses to make these assessments is protocol-dependent, and some communications protocols may employ a range of routing algorithms, and accompanying routing protocols. In the case of the TCP/IP protocol suite, the OfficeConnect Remote 840 utilizes RIP. RIP is also known as a distance vector protocol.
Different protocols use different networking characteristics or metrics when making routing decisions. The metric employed by RIP is a hop count. A hop count is defined by the number of routing nodes there are between the source and destination units. In our example, there are two hops between LAN 1 and LAN 2 going via Routers A and B. If traffic was directed via Routers A,D, and then B, this would be three hops.
The algorithm will automatically select to forward the data packet via Router A, as this route contains the least number of hop counts which makes it the preferred direct route.
Every thirty seconds (by default), each IP router will advertise, via RIP datagrams, to all other routers on the Internetwork, how many hops it takes to reach all connected logical networks, based on the routers network position and the state of its physical links.
It is also possible to assign what are known as static routes, which are manually entered fixed routes. The network manager may be aware of specific traffic patterns, or need to enforce a particular routing policy. Static routes provide an option to force traffic through the network in a particular way. The disadvantage with this approach is that routing protocols dynamically update all the routers on the network with the current network topology, enabling backup routes to be deployed. In a static route situation, if the WAN links in that routing definition are down, then traffic cannot be passed. Implementing a static route prohibits the router from being able to offer alternative data paths.