Setting Up and Accessing a Small Network - Walking the Walk: Topology and the Lay of the LAN

The Star Topology

The star topology, as you can see in Figure 1, consists of multiple workstations connected to a hub or router. In the most common scenario, each computer has a network adapter with an RJ-45 connector running a twisted-pair cable to a port in the hub. The hub (usually) just passes along the signals, so each computer gains access to the other computers on the network.

This is an excellent topology for peer-to-peer networks because it mirrors the no-machine-is-more-equal-than-any-other-machine philosophy of peer-to-peer. It’s also easy to add machines to the network because it’s a simple matter of running a new cable to the hub. If the hub’s ports are used up, you can connect a second hub to the first one. Another advantage of the star topology is that if one machine goes down for the count, the network access of the other machines isn’t affected. On the downside, star topology networks tend to need a lot of cable because you have to connect every node directly to the hub.

The Bus Topology

In a bus topology, shown in Figure 2, each node is attached to a single main cable called a bus or a backbone. For large networks, the backbone often extends throughout an entire building and is hidden behind the walls. For such lengthy cables, repeaters are often needed to boost the signal along various points of the backbone. Connections to the backbone are made via drop cables that run from network cards to wall jacks or some other junction box.

The big advantage of the bus topology is that it’s relatively easy to set up (aside from the effort required to run cable through a building’s walls), and its layout often mirrors the physical layout of an office or a building. The major drawback of the bus topology is that a break in the backbone brings down the entire network.

The Ring Topology

At first glance, the ring topology sounds suspiciously like the star topology. Each network node is connected to a central device, which is a special kind of hub called a multistation access unit (MAU), as shown in Figure 3. The difference lies in how the hub views the network. In the star topology, when the hub receives a packet, it checks the destination and then forwards the packet to the appropriate node without worrying about any other node on the network. In the ring topology, however, the circuitry of the MAU organizes the entire network as a ring, and each received packet is broadcast around the ring.

This is the topology used in token ring and ARCnet networks. This makes sense because these network architectures use a token-passing method to allocate cable access. The ring structure is a very efficient way to pass the token around to each node, so overall performance is improved.

Like star topologies, ring topologies are more stable than bus designs because one node going down doesn’t affect the entire network. There is one exception, however. The ring topology requires that each node actively pass along each packet. If a node goes down before it has had a chance to pass along a packet, the entire network crashes. Such a situation, however, is relatively rare.

Talking the Talk: Networking Protocols

In diplomacy, protocol defines the rules and formalities that ensure smooth communications between nations and cultures. A networking protocol performs a similar function: It’s a set of standards that define how information is exchanged between two systems across a network connection.

For example, consider what appears to be a simple procedure: exchanging a file between two networked computers. I mentioned earlier that files and all other network transmissions are broken down into packets. Because a large file might consist of hundreds or even thousands of packets, there has to be some way of coordinating how all this information is sent and received. Here are just a few of the questions that must be answered for even the simplest file transfer to succeed:

As you can see, it takes an incredible amount of give-and-take to coordinate any kind of network interaction. The inherent complexity of this process means that if the source and destination systems are even slightly out of sync, the file either will arrive corrupted or won’t arrive at all. Network protocols are designed to ensure that this doesn’t happen. The protocol specifies in no uncertain terms all the details of any kind of network transfer. Generally speaking, protocols fall into two categories: transport protocols and network protocols.

With a transport protocol (also called a connection-oriented protocol), a virtual communications channel is established between two nodes, and the protocol uses this channel to send packets between the nodes. Because the source and destination are defined in advance, the packets need not contain full address information. The constant link between the two nodes provides the protocol with an efficient path for exchanging messages, so this type of communications method is useful for applications that require a long-term connection (such as a network monitoring program). However, some overhead is involved in both setting up and closing the channel, so this method isn’t suitable for short-lived communications.

With a network protocol (also called a connectionless protocol), no communications channel is established between nodes. Instead, the protocol builds each packet with all the information required for the network to deliver each packet and for the destination node to reassemble the packets into the original file. These self-contained independent packets are called datagrams. The protocol then ships out the packets without notifying or negotiating with the destination node. All the network has to do is transmit the packets to the destination or to some intermediate stop along the way. This method requires a bit more packet overhead, but it’s efficient for short bursts because there’s no need to set up or shut down a channel between nodes.

Many protocols are available, but two are most common (these are the standard protocols available with Windows XP):

When setting up your network, you don’t have to commit to a single protocol. Windows XP is happy to work with multiple protocols simultaneously, so you don’t have to box yourself in. This is particularly handy in network environments that must access different types of machines. You can use IPX/SPX to access a NetWare server, and TCP/IP to access UNIX boxes, most Windows clients, and the Internet. However, most small networks operate best by using only TCP/IP.