This parameter is usually referred to as network cost. The cost of a hop can be increased by increasing the value a packet takes across a network path. The default preferred path is the low-cost path; the alternative path is the high-cost path. This arrangement prevents information from being routed over the high-cost path on a regular basis.

Figure 23.6 shows an additional router added between networks 1 and 3. The desired path for information from network 3 to network 2 is through server 1. Because router 2 connects network 3 and network 1, information can be routed between those two networks. Also, because router 2 is between network 1 and network 2, information is routed through that path. Information from network 3 that’s bound for network 2 can go over one of the two paths: either through server 1, or through router 2 and router 1. The latter isn’t the preferred path, because information can be routed directly over server 1. Therefore, a higher cost is assigned to the path of router 2 and router 1 from network 2. This type of path analysis must be performed in a multiple-segment network.

Choosing a Networking Configuration

The physical media used by an Internet network can be almost any network technology in current use. Internet network traffic isn’t limited to Ethernet, ARCnet, or Token Ring. It can travel over asynchronous RS-232, T1 lines, and through frame relay. Whatever networking topology is selected for the network, the configuration, installation, and operation rules associated with that networking technology must be followed.

Keep in mind the bandwidth that an application requires. Many applications require megabytes of data to be transferred, so bandwidth becomes a prime consideration. You can usually save bandwidth by compressing files before sending them over the network.

---

• See “Compressing Files,” p. 332

---

Another consideration is the physical location of the network. If all nodes are in the same building, a single LAN can be used. However, if the networks are located across town, a T1 connection may be needed. If the nodes are located in different geographic locations, a frame relay or a packet-switched network can be used.

In laying out a network, you must consider the type of information to be carried over the network, the physical location, and network loading. To help determine the capacity of the network, examine the type of workstations, servers, and applications.

If diskless workstations are used in a network, a higher network load is placed on the network for each node. The reason for this is that each remote diskless workstation requires all operating system code to be downloaded through the network. Because all applications, utilities, and data files are stored remotely, every action on that workstation requires network access.

Also of concern is the amount of NFS traffic that will occur on the network. NFS provides remote virtual disk services, so information retrieved and stored on these remote disks is constantly used on the network.

Other considerations are large graphical images, swapping and page files used for virtual memory, distributed database applications, printer traffic, and terminal traffic. These are all considerations in any network, but the designers and users of PC-based LANs don’t usually have to contend with them. When a network is connected into a general user community, all aspects of the networking environment come into play.

Other items to be examined are the need for dial-up and remote access. If this access is related to terminal and screen traffic, a serial port from an existing system may suffice. If a Point-to-Point Protocol (PPP) connection is made, you must consider how much overhead it will impose on the network when users are loading software utilities, programs, and databases over the phone lines. This is of concern because IP isn’t limited to a high-speed link such as Novell IPX and other networking protocols.

Understanding Network Configuration Guidelines

A network must be designed based on guidelines and rules. You should consider the following questions when planning a network:

•  How will the network be used today?

•  How will the network be used for the next several years?

•  What applications are going to be used on the network?

•  Will workgroups within the organization require networking resources in the future?

•  What types and numbers of workstations will be on the network?

•  How many servers, minicomputers, and other hosts will be on the network?

•  What other network devices, such as printers and plotters, will be on the network?

•  Will shared disk arrays and optical jukeboxes be necessary?

•  Will management of the network be centralized?

•  Will the network be connected to the Internet or other corporate networks, or will it be the basis for a wide area network?

•  What other protocols will use the networking technology (IPX, DECNET, LAT, OSI protocols, and TCP/IP)?

•  Where will critical data be interchanged (determine several different paths)?

•  How will the network grow and change?

After you address all these questions, the network can be defined. The number of nodes indicates how many class C address spaces are needed or whether a class B is needed.

Connection to remote facilities should also be addressed. The load can be distributed across multiple network segments. Try to minimize the traffic that has to go across networks. For example, if you have two systems that exchange a lot of information and hops across three networks are required for them to communicate, consider moving the systems to the same network.

Determine the best networking topology to meet the requirements specified in the network analysis. To allow for growth in the network, the best approach is to determine the maximum load and to develop a network in which that load is at a minimum.