"Linux Gazette...making Linux just a little more fun!"

Linux Installation Primer, Part 4

By Ron Jenkins


Copyright ® 1998 by Ron Jenkins. This work is provided on an "as is" basis. The author provides no warranty whatsoever, either express or implied, regarding the work, including warranties with respect to its merchantability or fitness for any particular purpose.


The author welcomes corrections and suggestions. He can be reached by electronic mail at rjenkins@qni.com, or at his personal homepage: http://www.qni.com/~rjenkins/. Corrections, as well as updated versions of all of the author's works may be found at the URL listed above.


NOTE: As you can see, I am moving to a new ISP. My old one changed to metered access, which makes the information superhighway a toll road. Please bear with me as I get everything in working order. The e-mail address is functional; the website will be operational hopefully around mid December or early January.




I would like to thank you all for your kind comments, and constructive criticisms concerning my articles. I would also like to thank the staff of the Linux Gazette, Marjorie in particular, for giving an unskilled goofball like me a chance to publish my scribbling. Keep those e-mails and questions coming!




To preclude a flood of e-mail on the subject, I have decided to change the order in which my columns will run. I had originally intended to do the IP_Masq/Internet Gateway piece this month, but then it occurred to me - what good is an Internet gateway without a network?


So, the new sequence for the next few months will be:


This column Planning a home network.

Deploying a home network.

IP_Masq/Internet Gateway.


If you can't wait that long, and have a need for the Internet Gateway stuff, just drop me an e-mail.


Part Five: Planning a Home Network

In this installment, we will address some of the issues necessary to plan a home network. We will cover most of the issues that you will encounter, and perhaps a few you had not thought of. Finally I will walk you through the steps to creating an effective and optimal Network Plan. As with each installment of this series, there will be some operations required by each distribution that may or may not be different in another. I will diverge from the generalized information when necessary, as always.


In this installment, I will cover the following topics:


  1. Do I need a home network or not?
  2. Some background theory on Ethernet and TCP/IP.
  3. Choosing a Topology.
  4. Choosing a NIC.
  5. IP issues - Reserved or Proper IP addresses.
  6. WAN connection issues.
  7. Planning the network - Physical vs. Logical layout.
  8. Planning ahead for easy administration.
  9. Deciding what services you require.
  10. Disaster Recovery and Fault Tolerance issues.
  11. Bringing it all together.
  12. References.
  13. Resources for further information.
  14. About the Author.


Do I need a home network or not?

This is a relatively easy question to answer. If you have more than one computer, you can certainly benefit by networking your boxes together. If you have a SOHO or small business, you can benefit as well.


You might ask, "Why do I need a network?"


Some possible answers include:

Integration of common services such as file sharing so that your documents are stored on a single machine, which in turn allows all or some of your users access.


Consolidation of all documents and data, eliminating the "Who's got the latest version of this freaking spreadsheet or document?"


The ability to create internal discussion forums, as well as access to newsgroups either in real time or off line relevant to your business or personal interests.


Consolidated Internet access for everyone where only one modem is required.


Fax and scanner access from all your workstations.


The desire to learn more about networking in general and Unix networking in particular providing you with new marketable skills.


Some background theory on Ethernet and TCP/IP.

For an overview of TCP/IP and networking, see my article in last month's issue.


Briefly, to network two or more computers, three things are required:

  1. A Network Interface Card (NIC) which is installed in the computer, and provides the physical as well as the logical (more on this later) connection to the network.
  2. A medium to exchange information from machine to machine. See Topology below.
  3. A common protocol to transport the data from machine to machine. In our case, TCP/IP.


Choosing a Topology.

Crucial to the proper performance of your network is the topology you choose. There are many different topologies available, but for the purpose of your installation, I will confine the choices to the two most common topologies - 10BASET and 10BASE2, or more appropriately a star network versus a bus network, respectively.


Pros and Con's of the two different topologies:



Uses unshielded twisted pair (UTP) wiring. Is a point to point topology, meaning if any node (computer) on the network goes down, the rest are unaffected.



Requires the use of a hub as a common connection point. Wiring is more difficult, since each node (computer) requires a separate connection to the central hub. More expensive than 10BASE2.




Uses easily available cheap coaxial cable forming a "bus" to connect all nodes. No hub or extra equipment required. Is easy and simple to wire. Costs significantly less than a 10BASET topology.



If the bus goes down, the entire network goes with it. Requires proper termination at both ends of the bus (basically two fancy 50-Ohm resistors). A termination problem can bring down the whole network.


Finally, another point to consider - mixed topologies are often used to accomplish different objectives. For instance, say you have an office set up in the basement that contains many workstations that are physically close together. Upstairs you have 3 computers used by your family in disparate locations. The solution - downstairs you use a star (10BASET) this provides better fault tolerance for your business machines. Upstairs you use a bus (10BASE2) to simplify wiring issues. To tie it all together, you run a 10BASE2 cable downstairs, extending the bus to the downstairs machines and hook it up to the hub. You can then access your "office" downstairs, to get your work done, and the business machines can contact you e-mailing you until they feel happy. Voila!



When determining the length of coaxial cable, remember that the cable will run from machine to machine, not in one long piece.


If you are going with UTP, depending on the size of your installation and amount of cable required, you may or may not want to look into purchasing the cable in bulk, purchasing some RJ-45 plugs, a crimping tool and do it your self.


Choosing a NIC.

This can be a tricky one. Almost everyone is tempted to buy the cheap clone cards, and sometimes it works, sometimes it does not. At least specifically ask if the card can disable the plug-n-pray features, as you may or may not need to explicitly set the IO address, as well as the IRQ.


This mostly applies to the ISA based cards. Most PCI cards can be autoprobed if you are using kernel 2.0.34>.


I like the 3Com products. They cost a little more, but it's worth it in the long run. For an ISA bus, I like the 509B. For a PCI bus, I like the 905 series. Also the PCI NE2000's are known to work. Also, the type of NIC you buy is largely determined by your topology choice. I recommend getting a "combo" card which contains both a 10BASET as well as a 10BASE2 interface. This lets you connect to either topology, and is a prudent measure.


As you will soon see networks are never a finished product, but rather a constantly changing, ever evolving project. Getting a combo card will give you maximum flexibility as your network changes. And it will.


A final note - NIC's are measured in the amount of bit space they can transfer data. Common to most Ethernet cards is 8, 16, and 32 bits. The higher the number the better. 8 and 16 bit cards are usually ISA cards. The 32 bit cards are PCI.


IP issues - Reserved or Proper IP addresses.

The next thing you will need to determine is the adressing scheme you will use on your network. I always tell my clients that getting Proper IP adresses (a block of IP's purchased from your ISP) is the best way to go, but it does cost more. This is usually referred to as a dedicated connection and costs more than a regular dialup account.


The advantages of a dedicated connection means your ISP will set aside one of their modems for your personal use. This, along with the IP addresses set aside for your personal use, account for the higher pricing.


Also, a dedicated connection allows you to have as many e-mail addresses as you want, put up your own website or sites, and for $74.00, your own domain on the Internet. This will give friends clients or browsers a permanent way of contacting you, obtaining information on your products or services, or a virtual gathering place for your family to let them keep in touch. As you and your family exchange more and more information, it can ultimately become the central point for family news, organizing events, and keeping current on things without those $50.00 phone calls everyone makes around Thanksgiving and Christmas.


More commonly, people want to used Reserved IP's - certain subnets, set aside to be used for this sort of service, and are not routable unless they pass through a gateway machine, or proxy, which effectively hides the interior network (usually 192.168.x.x) from the outside world making all your machines appear to the outside world as the gateway machine.


The downside to this is that using this scheme, you will only have one e-mail address, the one you got at the time of your sign up. However, many ISP's offer dialup accounts with more than one e-mail address, and some even allow concurrent connections (this means you can have more than one modem connected at the same time.) Check around in your area for this kind of service. It will probably cost more, but not as much as the dedicated connection option.


Finally, try to get a "static IP" address instead of a "dynamic" one. This will allow you to put up a webserver for personal use, or to advertise your business. Without a static IP, it is very difficult to do much more than pull from the Internet, you will not be able to push much more than e-mail.


Before I get bombed with e-mail about dynamic IP hacks, scripts that can post your current IP, etc. Please keep in mind that the purpose of this series is to provide new users of the Linux operating system as many services and options as possible, while keeping the configuration and deployment as easy as possible.


As the series progresses, and our skill levels improve, I will begin to go a little deeper into the details and tuning and tweaking.


WAN connection issues.

This is primarily a budgeting issue. Briefly you have two dialup choices, and for dedicated connections, you have three. Outlined below you will find the various choices compared and contrasted, along with my recommendations of what I usually choose.


Dialup Choices:

  1. A standard modem, 33.6Kbps or less. (What about 56k? I have not seen any so far that are not just a telephone interface and a impedance transformer, with all the "modem" work being done by your CPU. This is like P-N-P on steroids. If anyone has sucessfully used one, I would love to hear about it.) This is suitable for small networks, <=5 users, who will be using the Internet sporadically. This option costs the least. Requires a computer to function.
  2. An ISDN modem, and ISDN line. This option is best for networks of <=25 users, or power users who are on the net most all of the time, and doing many tasks simultaneously. I can and have soaked one of these all buy myself. But then again, I have nowhere to go and all day to get there. :-) This option will give you a true steady throughput of slightly less than 128Kbps. This option will require an additional ISDN line to be purchased, In my area, it runs $112.00 per month for unlimited time. There are metered usage plans that can run as low as $40.00 per month. This might make sense for you if you and your network will be sporadic users, but be warned - speed is addictive, and you may find your sporadic use goes way up. Additionally, your ISP connection charge may or may not be more. Requires a computer to function


Dedicated Choices:

Here you have both of the options above, and an additional one described below.


A dedicated router. This device takes care of the connection to your ISP, automatically redial if the link fails, and offers firewall and many other security features. It is an independent device, so no computer is required. All you need is the router and the ISDN line. Costs range from ~$100.00 - $800.00. I use the Ascend Pipeline 50, which as I recall cost about $600.00 when I bought it three years ago. This is the best choice for people with a dedicated connection, who plan to do business on the web as well as provide Internet access to their end users. Otherwise, it's probably overkill. This is the easiest, quickest, most reliable way to manage your connection. Can be set to dial on demand, from your network out, as well as from the Internet in. This may save you some money if you are on a metered usage plan. Your ISP charges will definitely be higher. In my area, a dedicated ISDN account ranges from ~$150.00 - $300.00 per month.


Planning the network - Physical vs. Logical layout.

There are two things to consider when planning a network the physical layout (where the machines are, where and how the cable will be installed, which machines will provide which services, etc.) And the Logical layout (how the data actually flows, and how each machine interacts with the network, usually expressed in a hierarchical manner.)


For instance, say you have a network consisting of four workstations, two on each side of another three machines, a fileserver, an Internet gateway, and a DNS server, all connected to each other by a bus (10BASE2) architecture.


Physically, you have 2 workstations, the file server, gateway, DNS, and two more work stations. Logically, you have four levels to your network - at the top you have your bus (since any interaction requires the bus to operate,) at the second tier, you have the Internet gateway and the DNS machines (since all machines require DNS to "find" each other, and DNS needs the gateway for name requests it cannot resolve,) at the third tier, you have the fileserver (since all the workstations need access to this machine, but it should not interact with the outside world for security reasons,) and finally at the fourth level, you have your workstations.


Planning both the physical and logical layout of your network is crucial to the effectiveness and performance of the network. On the physical side, you need to plan where your cabling will be, and pay particular attention to how it is placed. You will need to include in your plan entry and exit points if necessary and how you can best arrange the cables to run together and how you will bundle and anchor them. You will also need to consider the placement of any other network devices such as hubs or routers to keep the distance from the device to the machines that will connect to it to assure you will use the shortest length of cabling possible.


On the logical side, check and recheck your logical layout to make sure you are placing your machines in the proper logical positions that will provide maximum performance and minimum interaction problems. Looking at your network logically may point out some problems not apparent in the physical layout.


Planning ahead for easy administration.

Now we come to one of the two things most people do not or will not do, but are crucial to effective management of your network. You will need to do a thorough and complete inventory of all your hardware. At the bare minimum, you should collect the following information about every computer that will be connected to your network:


  1. Make, model number, and manufacturer of the computer.
  2. Type and speed of your CPU.
  3. Amount of RAM.
  4. Bus type.
  5. Number and type of slots used/available.
  6. The make, model, and manufacturer of each device inside your computer.
  7. The IO and IRQ for each of the above devices.
  8. Make, model, and manufacturer of you video card including the amount of RAM onboard.
  9. Make, model, and manufacturer of your monitor.
  10. What resolutions your monitor is capable of.
  11. Type and size of floppy drive(s).
  12. Type and size of hard disk drive(s).
  13. Type, make and model of your mouse.
  14. Make, model and manufacturer of any external devices.
  15. Type and version of operating system(s).
  16. Make, model, manufacturer and interface of your printer (if needed).
  17. Make, model, manufacturer and interface of your backup device (if needed).


Ideally, you should record everything, all the way down to the chipsets, but you can start with the above. I can hear everyone yelling "What good will this do me?"


Well, consider this - if your computer has only 4 MB RAM, and is running some flavor of windows, you will need to add more RAM. Similarly, if some of your workstations contain only ISA slots, while others have both PCI and ISA slots, now is the time to find out. Not after you get back from the store with a bunch of PCI NIC's.


The type and version of the operating system is very important. If you have any Novell boxes, they will require additional configuration and translation services. The same applies to some Mac's.


Additionally, this time and effort will pay off in the long run when, not if, one of your machines starts misbehaving.


Deciding what services you require.

This is important as well, because the services you need will somewhat dictate how your network is set up. Some of the more popular things are listed below. You may or may not have additional requirements.


  1. File Server - this will most likely be the first thing to think about. Consolidating access to your information was one of the reasons networks were invented.
  2. Internet access - this is the second most common service required. This will allow all workstations to connect to the Internet. Depending on the type of connection, you may or may not be able to e-mail, offer ftp services, and web services to the outside world, as well as internally. This will require either a router or a computer dedicated to this purpose. If you are using a computer to provide access, some additional configuration and software may be necessary.
  3. Name Resolution - some type of name resolution is required on any TCP/IP network. For smaller networks, you can simply use a hosts file to take care of this. If you have a dedicated connection, DNS is required. You must have two DNS machines to maintain your network information and when necessary, update the Internet root servers. Finally, if you are connecting through a dial up connection, you should probably consider running a caching nameserver from which all your network nodes obtain information, and in turn you instruct this machine to use your ISP's DNS servers. This will speed up things a bit on slower connections.
  4. If you are in an all Unix shop, or a cross platform environment, you will probably want to use NFS and possibly Samba. The former can be used by Unix machines by default, and on windows boxes with additional software. The latter is used exclusively by windows clients, making the Linux machine appear as just another computer in your Network Neighborhood, and allows you to transfer files by simply dragging and dropping, just like copying files from one disk to another.
  5. Sometimes it is advantageous to be able to execute programs on a remote machine, and have the results display on another workstation. Using telnet, you can execute any character mode programs, but often you will need and/or desire to run remote programs that require the X windowing system to function. Instructions for this can be found in the September issue of the Linux Gazette.
  6. Another handy thing to run is a time server. This allows all your machines to synchronize to the National Institute of Standards and Technology (NIST) atomic clock in Ft. Collins, Colorado. Many Internet applications and services are very sensitive to time disparities, and you want your servers to be right on, for examining the logs for problems or unauthorized use.

Disaster Recovery and Fault Tolerance issues.

I know I keep harping on this subject throughout my columns, but it is crucial. You WILL need a backup device. Ideally, you should have a backup device on every workstation and server on your network. Practically, you can get by with one backup device, usually on the file server, or a machine dedicated to this function.


When you purchase a backup device, make sure it is supported by Linux. Otherwise what you end up with is a very expensive bookend. This machine should have sufficient disk space to handle the spooling of your windows and Mac clients. Your Unix machines should be able to access the backup device remotely.


Also, you need to define a backup schedule for both your end users, as well as the servers. At a minimum, you should have enough tapes or whatever your backup device uses, to perform daily backups Mon. - Fri. as well a weekly backup Sat. or Sun. for two weeks. This will at least allow you to go back two weeks when, not if, you or one of your end users finds out they need a file they deleted "Uhh, sometime last week ."


Bringing it all together.

You have chosen your topology, picked your NIC's, decided on the type of IP addresses you will use, decided on the type and speed of your Internet connection (if needed,) looked at your proposed network from both a physical and logical point of view, completed your hardware and software inventory, determined what services you will require, last, developed a backup schedule and are going to purchase a backup device (if needed.)


"What do I now?"


You check everything over and over. You want to make all your mistakes at the planning stage, not the deployment stage.


Once you are satisfied with your plan, write it all down. What you need to purchase , as well as the things mentioned in this article. Then check it one more time.


Finally, you can start shopping around for the best price on the things you will need. Here are a few general guidelines - when purchasing coaxial cable, don't buy it at a computer store. The kind of cable they sell is crap and noisy as all getout. Go to a ham (amateur) radio shop, and tell them you want RG-58A/U coax with BNC connectors on each end in the lengths you require. If a Ham shop is not available, go to Radio Shack, and look there, where I believe they offer 6, 8, 12, and 50 foot lengths.


When purchasing your NIC's, look into bulk discounts. If you are buying at least four or five, there is often a price break.


Stay tuned, and next month we are going to actually install and configure the network !



The System Administrators Guide

The Network Administrator's Guide


The Ethernet HOW-TO

The IP_Masq mini HOW-TO

The Hardware HOW-TO


Resources for further information:




Previous ``Linux Installation Primer'' Columns

Linux Installation Primer #1, September 1998
Linux Installation Primer #2, October 1998
Linux Installation Primer #3, November 1998

Copyright © 1998, Ron Jenkins
Published in Issue 35 of Linux Gazette, December 1998