Network devices have one primary purpose: to pass network traffic from one segment to another. (There are exceptions, of course, such as network analyzers, which inspect traffic as it goes by.) With devices that independently make forwarding decisions, traffic can travel from its source to the destination. The higher up the Open System Interconnection (OSI) model a device operates, the deeper it looks into a packet to make a forwarding decision.
Railroad-switching stations provide a similar example. The switches enable a train to enter the appropriate tracks (path) that take it to its final destination. If the switches are set wrong, a train can end up traveling to the wrong destination or traveling in a circle.
Bridges and switches are networking devices that operate at OSI Layer 2. Bridges became popular in the 1980s and enabled packet forwarding between homogenous networks. More recently, bridges and switches forward frames among different types of networks.
Switching technology has emerged as the evolutionary replacement for bridging. Switches provide all the features of traditional bridging and more. Compared to bridges, switches provide superior throughput performance, higher port density, and lower per-port cost.
The different types of bridging include the following:
Transparent bridging primarily occurs in Ethernet networks.
Source-route bridging occurs in Token Ring networks.
Translational bridging occurs between different media. For example, a translational bridge might connect a Token Ring network to an Ethernet network.
Bridging and switching occur at the data link layer, which means bridges control data flow, provide transmission error handling, and enable access to physical media. Basic bridging is not complicated: A bridge or switch analyzes an incoming frame, determines where to forward the frame based on the packet’s content, and forwards the frame toward its destination. With transparent bridging, forwarding decisions happen one hop at a time. With source-route bridging, the frame contains a predetermined path to the destination.
Bridges and switches divide networks into smaller, self-contained units. Because only a portion of traffic is forwarded, bridging reduces the overall traffic devices see on each connected network. The bridge acts as a kind of firewall in that it prevents frame-level errors from propagating from one segment to another. Bridges also accommodate communication among more devices than are supported on a single segment or ring. Bridges and switches essentially extend the effective length of a LAN, permitting more workstations to communicate with each other within a single broadcast domain.
The primary difference between switches and bridges is that bridges segment a LAN into a few smaller segments. Switches, through their increased port density and speed, permit segmentation on a much larger scale. Modern-day switches have hundreds of ports per chassis. Additionally, modern-day switches interconnect LAN segments operating at different speeds.
Whereas switches and bridges operate at OSI Layer 2 (data link layer), routers primarily operate at OSI Layer 3 (network layer). Like bridging, the primary act of routing involves moving packets across a network from a source to a destination. The difference involves the information that is used to make the forwarding decisions. Routers make decisions based on network layer protocols such as Internet Protocol (IP) and Novell NetWare Internetwork Packet Exchange (IPX).
Routing gained popularity in the mid to late 1980s as a result of internetworks growing beyond the capability of bridges. Before this popularity, networks were relatively small and isolated, and bridges were able to handle the jobs of forwarding and segmentation. However, as networks grew, routers facilitated larger scaling and more intelligent growth across wider physical distances. Although routers are more expensive and complex than bridges, routing is the core of the Internet today. (As a side note, Cisco as a company made its name through routing.)
Routing involves two processes: determining optimal routing paths through a network and forwarding packets along those paths. Routing algorithms make the optimal path determination. As they determine routes, tables on the router store the information.
Routing algorithms fill routing tables with various types of information. The primary piece of information relevant to routing is the next hop. Next-hop associations tell a router that it can reach a particular destination by sending a packet to a particular router representing the next hop on the way to its final destination. When a router receives a packet, it attempts to associate the destination network address in the packet to an appropriate next hop in its routing table. In addition to next-hop associations, routers store other pertinent information in routing tables. For multiple paths to a destination, a routing table might contain information that allows the router to determine the desirability of one path over another.
Routers communicate with each other and maintain their routing tables through the exchange of messages over the network. Routing updates are one particular type of message. A routing update contains all or part of another router’s routing table and allows each router to build a detailed picture of the overall network topology.
Once a router determines an optimal path for a packet, it must forward the packet toward the destination. The process of a router moving a packet from its received port to the outgoing destination port is called switching. Although the process of switching a packet on a router is similar to that of a Layer 2 switch, the decision criteria and the actual handling of the packet are different.
When a computer determines that it must send a packet to another host, it places the network address of the final destination host in the packet. However, it places the Layer 2 physical (Media Access Control [MAC]) address of the nearest router in the packet. When the router receives the packet, it first determines whether it knows how to reach the packet’s stated destination network. If the destination is not known, the router typically drops the packet. If the destination is known, the router changes the destination physical address in the packet to contain that of the next hop. The router then transmits the packet out the destination interface.
The next hop can be either the final destination or another router. Each router in the process performs the same operation. As the packet moves through the network, each router modifies the physical address stored in the packet but leaves the network address untouched (because it determines the final destination).
In an ideal world, each thing does what it is defined to do. This is not the case for network devices. Routers can provide bridging functionality, and switches are quickly becoming the high-density port, high-speed router of the campus. Network devices, including switches and routers, make forwarding decisions on OSI layers higher than the network layer. For example, routers can provide firewall functionality in which the router inspects Layer 4 packet information, and switches such as content switches can perform forwarding decisions based on Layer 5–7 packet information (such as the URL in an HTTP packet).
Flat Versus Hierarchical, Intradomain Versus Interdomain
With flat networks, all routers must keep track of all other routers on the network. As networks grow, the amount of information contained in the routing tables increases.
Although this method is simple, it can result in poor network performance because the number of routing updates traffic grows with each new router.
Hierarchical networks segment routers into logical groupings. This arrangement simplifies routing tables and greatly reduces overhead traffic.
</division><division> <title>Intradomain Versus Interdomain</title>
You can easily understand intradomain and interdomain routing in the context of large-scale hierarchical networks. Think of each segment as its own autonomous network. Within each autonomous network, intradomain routing protocols, also called Interior Gateway Prototols (IGPs), exchange routing information and forward packets. Interdomain routing protocols, also called Exterior Gateway Protocols (EGPs), are used between autonomous networks.
</division>At-A-Glance—Distance-Vector and Link-State Routing
The two main classes of routing are distant vector routing and link-state routing.
With distance-vector routing, routers share their routing table information with each other. Also referred to as “routing by rumor,” each router provides and receives updates from its direct neighbor. In the following figure, Router B shares information with Routers A and C. Router C shares routing information with Routers B and D. A distance vector describes the direction (port) and the distance (number of hops or other metric) to some other router. When a router receives information from another router, it increments whatever metric it is using. This process is called distance accumulation. Routers using this method know the distance between any two points in the network, but they do not know the exact topology of an internetwork.
</division><division> <title>Discovering Information with Distance Vectors</title>
Network discovery is the process of learning about indirectly connected routers. During network discovery, routers accumulate metrics and learn the best paths to various destinations in the network. In the example, each directly connected network has a distance of 0. Router A learns about other networks based on information it receives from Router B. Routers increment the distance metric for any route learned by an adjacent router. In other words, any distance information A learns about other routers from B, it increments by 1.
</division><division> <title>Link-State Routing</title>
With link-state routing, also known as shortest path first (SPF), each router maintains a database of topology information for the entire network.
Link-state routing provides better scaling than distance-vector routing because it only sends updates when there is a change in the network, and then it only sends information specific to the change that occurred. Distance vector uses regular updates and sends the whole routing table every time. Link-state routing also uses a hierarchical model, limiting the scope of route changes that occur.
</division>