Installing a Small Office Network

Planning and Wiring Basics

When a small office decides to move from a serial‑only environment to a true Ethernet network, the first thing that surfaces is the wiring plan. Serial links feel simple because a single cable can tie two terminals together, but they don't support the kind of bandwidth or flexibility that a modern office requires. Ethernet brings the promise of 10 Mbps - or 100 Mbps with the right gear - alongside the ability to grow the network as the company expands.

There are two main wired options that still make sense in a budget‑conscious setup: 10BASE‑T and 100BASE‑TX. 10BASE‑T uses twisted‑pair cabling, which is inexpensive and easy to run, while 100BASE‑TX requires the same cable but offers ten times the speed. 10BASE‑R and other older coax‑based standards like 10BASE‑2 are now historical footnotes; they lack the ease of installation and the reliability of modern twisted‑pair.

At the heart of a 10BASE‑T network is the hub or, more commonly today, the switch. A hub simply repeats whatever data it receives to every port, so a break in one cable can bring down the whole segment. A switch, on the other hand, forwards traffic only to the intended destination. If you use a 10 Mbps hub, the cost is about $20–$30 for a 4‑ or 8‑port model, which makes the trade‑off obvious for most offices that can spare the money. Switches also reduce congestion and eliminate the need for terminating resistors that coax‑based networks rely on.

When planning cable runs, keep the 328‑foot (100‑meter) limit in mind. Exceeding this limit forces you to add a repeater or switch, which adds cost and potential points of failure. A single switch located in the center of a floor works well for a cluster of 10–15 workstations, but if you need to cover multiple rooms or a two‑story layout, a hierarchical design with two or three switches connected by high‑speed uplink cables will keep traffic moving smoothly.

Cable choice is crucial. Ordinary telephone wire can handle data but does not guarantee the signal quality that a busy Ethernet link demands. Category 5 (Cat‑5) cable is inexpensive, has a sufficient twist rate to reduce crosstalk, and is widely available in 10‑foot and 50‑foot lengths. Buy Cat‑5 that’s rated for 100 Mbps if you want to future‑proof the office. Each cable ends in an RJ45 connector, and the cable itself is typically wrapped in a protective sleeve so you can run it through walls without damage.

The next step is terminating the cable. Rather than crimping RJ45 plugs directly onto the cable - a tedious process that still leaves room for errors - you’ll typically install wall jacks that house the cable and provide a convenient point for patch cords. The wall jack will have a small faceplate and an interior board that the twisted pair attaches to. A pair of pliers is all you need to strip the cable, arrange the wires in the correct T‑568A or T‑568B order, and secure them in the jack. Once the wall jacks are in place, you run patch cords - short, pre‑terminated cables - to each workstation. The patch cords keep the network organized and let you move or replace equipment without re‑terminating the cable.

While it is tempting to do all of this yourself, the wiring job can get messy. A professional installer knows how to follow building codes for running cables through walls and ceilings, and can help you avoid issues like signal attenuation or interference from electrical lines. If you decide to take the DIY route, double‑check that you’re using the correct cable type, that all connectors are firmly seated, and that you’ve marked each patch cord so you know which port belongs where.

Once the physical layer is complete, the next phase involves selecting the right network interface cards (NICs) and installing the software that will make your office machines talk to each other.

Choosing and Setting Up Network Hardware

With the cabling done, you now need to equip every workstation with a NIC that supports the speed you’ve chosen. For 10 Mbps, any inexpensive PCI or PCI‑Express card will do. For 100 Mbps, look for a card that explicitly lists support for 100BASE‑TX and has a driver available for your operating system.

Windows machines make this step painless. A new NIC usually comes with a CD that contains the driver and an installation wizard. Once you plug the card into the computer, Windows will detect it, install the driver, and you’ll see a new network adapter appear in the Network Connections folder. From there, you can enable TCP/IP, set a static IP address, or point the computer to a DHCP server that will hand out addresses automatically.

SCO Unix and Linux systems, however, demand a bit more care. Before you buy, check the vendor’s hardware compatibility list to ensure drivers are available. On SCO, you add the card via Hardware->Add New Lan Adaptor, then enable the TCP/IP protocol under the adapter’s properties. Linux users can use ifconfig or ip addr add to bring the interface up and assign an address. If the card doesn’t appear automatically, try adding any dummy adapter first; this trick often forces the system to detect hidden PCI devices.

Drivers themselves can make or break performance. Cheap cards sometimes lack the full set of features or have buggy drivers that lead to packet loss or dropped connections. For servers or machines that need high reliability, it pays to spend a little extra on a card with a proven track record in the Linux or SCO community.

Once every machine has a NIC, you need to connect them to the network device. For a hub, you simply plug a patch cord into each port on the hub. For a switch, you connect the uplink ports between switches if you’re using a multi‑layer design. The switch’s LEDs will flash as traffic passes through, giving you an instant visual check that the cables are functional.

Now it’s time to configure IP addressing. If you’re running a small office of less than 20 machines, manual static assignment works fine. Pick a private network range - 10.0.0.0/8, 172.16.0.0/12, or 192.168.0.0/16 - and hand out addresses that reflect the location of each device. A simple convention is to use the port number on the patch panel as the last octet: for the device on port 1, for port 2, and so on. Keep the router or gateway at the end of the network on or , whichever you prefer. This scheme makes troubleshooting easy because you always know which device occupies which address.

For a growing network, or if you want to simplify management, deploy a DHCP server. The server maintains a pool of addresses and hands them out on demand. It also provides the subnet mask, default gateway, and DNS server information. On Windows, DHCP can run as a service on a dedicated server or even on a workstation that stays powered on all the time. On SCO Unix, the dhcpd daemon can serve the same purpose. Once DHCP is in place, each workstation just needs to enable “obtain an IP address automatically” and will receive all the necessary network configuration from the server.

Regardless of whether you use static or dynamic addressing, make sure you document the IP plan. A simple spreadsheet with hostnames, IPs, and MAC addresses saved in a central location will save you headaches during maintenance or when you add new devices. Also remember to reserve a few addresses for future expansion and for network services such as a print server or a backup appliance.

With hardware and addressing settled, you’re ready to think about how the office connects to the wider world. That step introduces security considerations and the choice between NAT, a proxy, or a dedicated firewall.

IP Addressing, Security, and Internet Access

Once the local network is up and running, the next milestone is connecting to the Internet. Most small offices achieve this through a single broadband modem - often a cable or DSL line - that the server talks to via PPP. The server then acts as the gateway for all other machines. The simplest scenario is to run a basic mail server that pulls and sends email via the ISP’s SMTP/POP servers. If you need broader Internet access, you’ll need to decide between NAT, a proxy server, or both.

Network Address Translation (NAT) is the most common approach. A NAT device translates the private IP addresses of the office to one or a few public IP addresses that belong to the ISP. When a client sends a packet to the Internet, the NAT router changes the source address to its public address, stores the mapping, and forwards the packet. Replies come back to the public address and the router looks up the mapping to deliver them to the correct internal host. The advantage is that one public IP can serve dozens of devices, and the internal network remains hidden from the outside world. Most modern routers bundle NAT with DHCP and basic firewall rules, making setup a matter of entering the ISP’s credentials and pressing “apply.”

Software NAT is also available on SCO Unix and Linux. The ipfilter tool on SCO or the ipchains command on Linux can perform NAT with a configuration file that specifies which subnets are allowed to reach the Internet. For example, a SCO command might look like ipnat -f - 0/32, while a Linux rule could be ipchains -a forward -s 192.168.1.0/24 -j MASQ. These tools also double as basic packet filters, giving you a first line of defense against unwanted traffic.

Proxy servers sit one layer above NAT. Instead of translating addresses, a proxy server receives HTTP, HTTPS, FTP, or other requests from internal clients, fetches the content from the Internet, and returns it. Proxies can cache frequently requested data, reducing bandwidth usage and speeding up response times for multiple users. Popular open‑source options include