Fundamentals of an Internetwork
01/30/2001Networking 101A
Back in the days when I went to college, some people selected their academic training based on whether they would have to use a computer or not. My wife, having an aversion to computers, went into the field of graphic arts because at that time it was an art done by hand. Things have changed considerably. My wife's trade is now nearly paperless, and my nuptial duties have been expanded to include 7x24 IT service to her home office.
The Internet revolution has redefined the means by which we do business. It doesn't matter if you are a small business owner, the CEO of a large corporation, a teacher, a travel agent, stock broker, realtor, or manage a church office. The Internet has become an unstoppable phenomenon that's forcing seasoned professionals back into the classroom.
This is the first of a series of articles, aptly titled "Networking as a Second Language," that will cover a broad scope of networking aspects. We'll start out with real basic terminology and then move to advanced concepts such as configuring network hardware. It's all here for you. Let's get started.
Two computers and a wire
Since we're learning from the ground up, let's start out with some basic definitions. Fundamentally, what exactly is a computer network? To answer this I can provide you with a real-world example from my ancient past.
In the late 80s, I worked with Apollo Unix workstations. The Unix operating system allowed other Unix machines, called hosts, to access the hard drives across a network. To accomplish this, two or more systems were connected together using coaxial cable, very similar to the coax used for your cable television hookup. The Apollo workstations were chained together with this single piece of coax as it snaked through the work areas of each workstation connected to the cable.
If another workstation wanted to tap into the network, one would merely go down to Radio Shack and pick up a BNC T-connector to hook up. This network in its crude form was the basis of many early office networks. The early 90s PCs used the same cabling and special software called ArcNet, LanTastic, and Novell, to name a few.
So what is a computer network, you ask? In our crude example, a computer network is comprised of two or more computers, called hosts, communicating over a wire. As you can see, early computer networks basically were hosts attached to a wire.
Digital etiquette: bursts, bits, and packets
We have established that a computer network requires two hosts and a wire. Now we must examine how the hosts use this wire to communicate.
A common phrase today is that we live in a digital age. The term "digital" means that a device uses bursts of electricity in discrete values. That is, a burst may be sent out over the wire of five volts. This one discrete burst would represent a bit set to one. A low voltage burst of zero represents a bit set to zero. A bit belongs to a binary number system whose values consist of 0s and 1s. The binary system has only those two digits. Hence the name binary digits or simply, bits.
When a host communicates on the wire, it may send, as an arbitrary example, a sequence of 48 bits at one time onto the wire. This sequence of 48 bits is referred to as a packet. A packet is a discrete number of bursts onto the wire whose contents are understood by the hosts connected on the network.
Computers use this etiquette when speaking to one another. This is called a protocol. Network protocols are the subject of a future installment. It's presented here to give you an idea of the behavior of hosts and how they communicate on a shared wire (media).
The sequence of bursts onto the wire (the packet) contains information on where this packet originated and which host it is destined for, as well as data. This exchanging of packets between computers on a network is referred to as a packet-switched network. Although this term isn't commonly used outside of academic lectures, this concept of breaking up of data into pieces, packets, is the fundamental concept of network operation.
Network architectures
The simple network we have been examining with multiple hosts on a network is called a local-area network, commonly called a LAN. A LAN is not limited to a few hosts and one network segment. A LAN typically can range in size to hundreds or even thousands of hosts on multiple networks. The breadth of a LAN can be limited in size to one building or as large as a corporate campus. The key point of a LAN is that the wiring for the network is centralized to the location of the network.
The LAN I mentioned earlier with the coax cable became unmanageable. Black coax cable was snaking all over the office space and through cubicles as well as dropping down from ceiling tiles. The next migration for a company was to install a wiring closet, commonly referred to today as a data closet.
The data closet allows all network cabling to be centralized in one room or several rooms (depending on the size of the institution). The network wiring is patched into a network hub. Likewise, network wiring is pre-wired to a cubicle or office space and runs back to the data closet. Hooking up a host to the network is now a matter of patching the cubicle connection into the network hub. This is a much cleaner approach to the snaking coax cable.
Our network knowledge has now expanded to LANs and hubs in one centralized site. So what happens when a company grows and needs to connect LANs, separated physically for business reasons or by geography? We now have two separate LANs, referred to as LAN segments, that we want to connect into an internetwork.
In the late 80s I was involved with one such setup. A startup hard-drive company was growing in size and acquired office space down the road to accommodate the prosperous growth. The requirement for this network was to build a bridge between the current LAN segment and the new office space. This was accomplished by working with Pac Bell and leasing two T1 56K lines between the two sites. A network device called a bridge sat physically in each building at the termination of the T1 connections.
As business for the hard-drive company continued to prosper, more employees were hired to meet its growing demands. More buildings were leased and built to house the new employees. This meant more LAN segmentations and a greater demand for user connectivity to the network.
LAN bridges were architecture solutions in the early part of the company's growth. Soon the flat, bridge-based network was becoming unmanageable. Bridges merely repeat traffic from one end to the other. Some segments such as engineering and accounting wanted to conceal their traffic for bandwidth and security reasons. Hence, the flat-earth network of bridges migrated to an environment with network routers.
Network routers allow for scalability and a mixture of media. That is, a network using wire can be connected to a network of fiber optics, such as a FDDI (pronounced fiddie, an acronym for Fiber Distributed Data Interface). Such a network would use the FDDI ring as a backbone to the corporate network. FDDI rings, which are still in use today, have a bandwidth of 100 megabits per second (Mps). In this tiered network architecture, multiple routers attach to the FDDI backbone. Typically a router in the wiring closet of each building on the campus would connect to the FDDI.
Newer technologies have emerged, such as the optical router, gigabit router, and routing LAN switches. However, the basics really haven't changed, especially in corporate network markets. Network infrastructures are costly to remove, so the FDDI backbone I described is a legacy and predominant architecture that's still in use today.
Feeding the desire to learn more
By now you might be anxious to learn more about networking. You can sign up for a CCIE course that runs well over $1,000. There are some good sites on the Internet such as routergod.com for entertaining insights on learning network architecture and design.
A book that I often recommend is "TCP/IP Illustrated, Volume 1: The Protocols," by Richard Stevens. This is the classic text that is found on every network engineer's bookshelf.
Of course, I also want to mention that the next installment of Networking as a Second Language, where I'll discuss the network layers, should be immensely helpful. Until then...
Michael J. Norton is a software engineer at Cisco Systems.
Discuss this article in the O'Reilly Network Forum.
Return to the O'Reilly Network Hub.
