This month's article is written for people who are just getting to know computer networking, or for those who want to increase their knowledge of the basics.
Computer networks were invented to connect computers together to allow them to share resources such as files and printers. Of course, networks have been around for quite some time. Early network implementations were proprietary (because there was nothing else). The networking capability was written into the application (an accounting system for example), and the choice of network protocol (communications software) and hardware was hard-coded into the application package. But this created problems. If the user or manufacturer wanted to change any of the networking components, the core application had to be rewritten. When networking really took off, it seemed that there was a “protocol flavor-of-the-month.” Things were changing so fast that it became impractical for application vendors to keep up. There had to be a better way.
In response to this problem, engineers developed the open systems interconnect (OSI) layer model for networking. This layered model allows the application (an accounting program, editing application, etc.) to remain separate from the layers below. This way, software protocols and hardware can be upgraded without having to rewrite the overarching applications. The application then interacts with the protocols through a software interface called an application programming interface (API). Standardization of the API allowed manufacturers to substitute different technology at lower layers without having to overhaul their applications.
Figure 1. Simplified layered model. Click here to see an enlarged diagram.
The layered model in Figure 1 illustrates another fundamental concept in networking. Note that the protocols are a separate layer from the physical layer. This allows a given networking technology such as Ethernet to be implemented over different physical media. For example, users could implement Ethernet using unshielded twisted pair (UTP) or optical fiber. In fact, these different physical media types could be mixed within a network. The signals riding on these different media types are Ethernet. The layered approach to networking makes this possible.
It also is possible to select protocols independently from the physical hardware upon which these protocols are transported. For example, both Ethernet and Fibre Channel can be carried on fiber.
There are many variations of the OSI model. The most common has seven layers: physical, data link, network, transport, session, presentation and application.
While it is not important to understand what all these layers do, you should know that there are fundamental differences in how network signals are moved, and that many of these differences center around whether switching decisions are made at Layer 2 or Layer 3. It is not always possible to make a clear distinction between the different layers. There are many excellent OSI model tutorials available on the Internet.
The two dominant networking technologies these days are Ethernet and Fibre Channel. Ethernet, a packet-based networking technology, has by far the largest technological deployment. Data from an application is sent to the protocol layers. In these layers, the data is chopped up into packets. Next, a destination address is put on the front of the packet, and the packet is sent to the physical layer for transportation on the network.
Figure 2. Simplified Ethernet packet. Click here to see an enlarged diagram.
Figure 2 shows a simplified Ethernet packet. Thousands of these potential Ethernet packets are generated by a computer each second and shipped across the network, each packet traveling independently. The address on the packet allows the network to route the packet to its destination. The protocol layer at the destination computer is responsible for reassembling the packets in the proper order and presenting the application with the original data.
Ethernet is used ubiquitously both for intranets (networks that are local to a given facility), and on the Internet. The Internet (capital “I”) is comprised of a large number of networks and switching technology that allows computers to send data across the country or around the world. While the switching and scope of these networks is vastly different from that of an intranet, the basic Ethernet packet remains unchanged. In fact, usually the same Ethernet packet travels across a local intranet, through a gateway computer and on to the Internet.
Figure 3. Hub-and-spoke Ethernet network. Click here to see an enlarged diagram.
There are many different Ethernet topologies. Topology refers to how computers in the network are connected together. The most common topology is called hub-and-spoke, in which each computer has a single, dedicated Ethernet connection to a central Ethernet switch. (See Figure 3.) If Computer A is transferring files to Computer B, it can do this at full speed while Computer C is transferring files to Computer D.
A star network is easy to build and troubleshoot, and it can provide high bandwidth to the desktop if it is designed properly. But there are caveats. To get the maximum bandwidth between devices connected to the network, the switch itself must have the capacity to operate at double the bandwidth of the individual connections to the computers. In our example, if this is a 100Mb/s network, the switch must have at least 200Mb/s of available bandwidth.
Fibre Channel is a computer network that is frequently confused with Ethernet. They are two separate, and incompatible technologies created to solve different problems. The confusion arises because they both can run on the same physical network. It looks as if you can just plug a Fibre Channel cable into an Ethernet fiber switch. But this will not work. The two networks use fundamentally different protocols or language to talk, and they come from different origins. Fibre Channel was created to connect computers to disk drives. In the early days of computing, there were strict limits on how far the disk drive could be physically located from the computer itself. Remember, CPUs sat in one box, and storage sat in another box. As computers got faster, they needed faster connections to the disk drives that served them. Parallel connections to drives became the norm. But this too reached a practical limit as the lengths of parallel cables started to give rise to termination problems, RF crosstalk, and poor frequency response. Computer designers needed a cable extender for disk drives that could be easily supported on existing systems. At this time, the Small Computer System Interface, or SCSI, was being used for many high-performance drives. Network engineers went to work designing a computer network that could transport SCSI commands. They came up with Fibre Channel — a network that established virtual connections between devices, allowing the robust transmission of SCSI commands across virtually unlimited distances. Now, Fibre Channel is the predominant local network in use for the connection of high-speed peripherals to computers.
There are a number of other networking technologies in use today, including Firewire and wireless. These networks are usually variations on the theme of packet-based interconnections created to allow the transfer of data from one computer system to another. If you do not know where to start, you might consider enrolling in a basic networking course at a local community college. You might also consider purchasing a couple of inexpensive computers and a switch and see what you can do. There is nothing like hands-on experience.
Brad Gilmer is president of Gilmer & Associates, executive director of the AAF Association and executive director of the Video Services Forum.