Read an Excerpt
Chapter 2: The Campus NetworkThrough the evolution of computers and networking, changes are occurring at a dizzying pace. This is apparent in the roles that hubs, bridges, routers, and switches are playing in small company and large corporate networks. Today, switches are at the center of networks, performing vital tasks in allowing users to carry on with their day-to-day activities. However, this was not always so. Traditional networks posed varying problems for network administrators, and they had repeaters, bridges, and routers to help provide solutions to these problems.
Networking technology and its related equipment has become very complex. Switches have replaced bridges, and Layer 3 and Layer 4 switches are beginning to replace routers. Today's corporate networks are being designed to carry different data types, such as voice, video, and data traffic, with technologies such as Gigabit Ethernet and Asynchronous Transfer Mode (ATM), and the necessities of quality of service (QOS). So it's not too surprising that many network engineers are scratching their heads with all the solutions that vendors bring to their doorsteps. This chapter shows some of the scalability problems that your campus intranet will face when trying to meet the everincreasing demands of your business. The success of your network will be based on the type of design you implement. A hierarchical design with the correct placement of your network services will provide your campus network with successful scalability.
Traditional Campus Networks And Their Issues
A campus network, or intranet, can be described as many different things. It can be a group of interconnected LANs, the network in a building, or the networks connected between different buildings. Cisco's definition of a campus network, or intranet, is defined as a group of interconnected LANs that are in the same geographic location and maintained by one group of people. Intranets can include such LAN technologies as Ethernet, Fast Ethernet, Gigabit Ethernet, Fiber Distributed Data Interface (FDDI), Token Ring, and even ATM.
From the early 1980s to today, the explosive growth of local area networks has been phenomenal. Traditional LANs of the '80s and early '90s and their support and maintenance were fairly straightforward compared to today's available technology solutions. At the network administrator's disposal were varying devices, such as hubs or repeaters, bridges, and routers, to solve problems in their networks. However, each of these devices only solves certain problems and sometimes at a great cost.
Initially, LANs started as a simple phenomenon: a handful of computers that were somehow interconnected. If the network administrator was using Ethernet, he might have used 10Base2 or 10Base5, thinnet or thicknet coaxial cable, and maybe some Ethernet hubs. Some of these devices might have been PCs, a file server or two, and maybe even a minicomputer. As his network grew in numbers and over a greater distance, he started facing certain problems.
Adding users to this network scheme is quite simple, but it does present problems. Because traditional network topologies, such as Ethernet, are shared environments, the more users that are placed on a segment, the less bandwidth each user will receive. The users share the bandwidth on a segment-for good or bad. Because of the demands a new user will have for resources on the network, it's sometimes hard to predict what impact this will have on a given segment of the network or on the network itself. Sometimes the network might be fast, and other times it might slow to a crawl, depending on what the users are actually doing. Because of the explosion of the Internet, it's not uncommon for users to download large text and image files, which can cause havoc even on a well-behaved network.
Besides having users on a shared medium, there are other problems that adding new users introduces. If the network consists of an Ethernet topology, more collisions will occur as more users are added, further reducing available bandwidth for their networked applications. Ethernet works under the premise of Carrier Sense, Multiple Access, Collision Detection (CSMA/CD). The "Multiple Access" part describes the shared environment of Ethernet itself-there are many stations sharing the bandwidth on a single piece of wire (or a set of wires interconnected via a hub). Since only one station can send information at a time, this is where the "Carrier Sense" part comes in. A station checks the wire to verify that no frame is currently traversing the wire before putting its own frame on the wire. Just in case two stations check the wire simultaneously and see no traffic on the wire and respectively place both of their frames on the wire, causing a collision, there is a detection mechanism built into Ethernet to verify the valid transmission of a frame. Collisions are not necessarily a bad thing-they're just a part of how Ethernet functions. However, the more users access resources on the network, the more likely that collisions will occur. Collisions become a concern when they start having an impact on the users on a segment-the network slows down, causing complaints from the users. If a Network Interface Card (NIC) card goes bad or a cabling problem occurs, inadvertent collisions occur, sometimes creating havoc for all the users on that segment.
Another problem that's equally as important as collisions is the issue of broadcasts. Most LAN operating systems such as Microsoft's Windows NT, Novell's NetWare, and Apple's AppleTalk protocols make use of broadcasts and/or multicasts to help users easily find resources on the network. Of course, the problem with this is that it further reduces the bandwidth available on a segment and has a negative impact on the performance of the computers on that segment. There are three basic types of broadcasts:
A unicast is a frame that only a specific computer will actually process. When a Layer 2 Ethernet or Token Ring frame is placed on a wire, every computer on that segment sees the frame.
Exam Alert: With a unicast frame, the destination MAC address field has a specified computer's MAC address. Every NIC card of every computer on the segment sees it; however, only the actual destination computer's NIC card will recognize the address, send an interrupt to the CPU, and send the frame up the protocol stack for further processing.
If the MAC address in the destination field does not match its own, the NIC card will discard the frame.