Jack Scott's Blog

I’ve spent a lot of time creating and fixing computer networks, both professionally (as a computer technician for a local computer shop) and in my private life (long time readers will know of my affinity for acquiring servers). So I thought it might be a good idea (prompted by a suggestion from one of my friends Michael) to give a few tips to those just starting managing small networks. Note: I have no real experience with medium to large-size networks. Some of my advice may still apply (and I have done a bit of research), some of it won’t. I’ll leave it up to the experts in those fields. But here are my tips on physically connecting computers in a small network.

Colour-coded Cables

This is the really big, obvious, thing to do. Whether you colour code for uses (blue for main cables runs, red for patch cables, etc) or for type (blue for Cat5e, red for Cat6, beige for telephone, orange for fibre, etc) doesn’t matter, as long as the cables all follow the same pattern, and you (and anybody else likely to work on the network) can understand the code. I myself code by type, as I believe it will be useful in allowing me to see if I need to upgrade cable in the future. I *could* go through and read the writing on all the cables, but in a darkly-lit room or under-floor cavity, this is a lot harder than looking for a colour.

It pays to remember that most network cable that comes in large rolls is blue Cat5e or Cat6, and to allow for this when designing a colour code system.

Labelling things

Once a network gets past a certain size (of about 2 or 3 network devices, not including client machines) it becomes important to have a good labelling system in place. It doesn’t really matter what this system consists of. I used to use masking tape and a felt-tip pen, and I now use a Dymo labelwriter (not purchased specifically for the job, but it does it brilliantly). As for what I label and what the label says:

• On network cables (of all sorts) I generally label each end with where the other end connects. So on the switch end of network cables, I write which device the cable connects, usually using a hostname (like “router” or “lounge-pc”).
• On network devices (such as routers) I usually label the device with its IP address, and if it is unlikely to be seen by visitors, also the password to access the device. Writing down the password on the device may seem like a bad idea, and can be in some situations, but it does save headaches. For most home or small office networks, the most likely threats to network security are likely to be hackers on the Internet, not those inside the building with access to the devices.
• On client machines (and servers) I write the hostname and the IP address (if statically assigned) on the front of the machine. Most people are unlikely to have enough computers that they can’t remember which one is which, but occasionally it does happen, especially when developers such as those writing operating systems need test beds for as many pieces of hardware as they can.

Naming Machines

Give all machines a decent, understandable name. If you’re only going to have one of something in your house, name it after the function. I used to have ‘server’, ‘lounge-pc’, ‘jack-laptop’ and so on in my house. If you’re going to have more than one server or one lounge PC, there are a couple of ways of going about it:

• Pick random names following a theme. I’m now using planet names for servers and moon names for desktops, and my network router is called Sol.
• Build up a name for the machine using a standard pattern. In this case, ‘pc273-1′ might mean the 1st PC in room 273. It’s more useful for large numbers of computers.
• Something else entirely.

This isn’t really related to the physical network, but it does have an influence, and is important to get right. Changing it can be a bitch for larger numbers of computers.

Get a rack!

Once you get past two or three routers, modems or access points (or whatever it is you have than doesn’t stack well), in my opinion it’s time to start looking at rackmounting some of the equipment. Face it: if you have this much stuff, it’s unlikely that you’re ever going to have less of it. So best figure out a way to keep it all neat and tidy. A solution to this problem has been available to the masses for a few years now, thanks to the wonders of eBay.

It’s possible to buy the rackstrip and bolts you will need online, and then build the enclosure yourself. For all the cases where you only need a single rack, I’d recommend going this way. You’ll end up with a custom solution that the right size for your needs, and it’ll end up cheaper (trust me). Just remember to add a bit more space than you need now, for any future expansion. It’s not cheaper if you have to build it anew every 3 months.

A note: it is possible to make the rackstrip yourself as well, but unless you are very skilled with metal and drill presses, I wouldn’t bother. It’s too much of a time investment.

A rack allows you to keep all your networking equipment in one place, and hide it out of the way. You can even buy shelves to put non-rackmount equipment on. If you want to rackmount your servers too, go for it. A decent 2U or 4U ATX rackmount case is not a lot dearer than a decent tower ATX case. More expensive, granted, but not by a lot once you consider quality.

Doing things properly

If you’re reading this article, chances are you’re interested in networking. And if you’re interested, make it enjoyable. To make it enjoyable, do things properly. Why? Because doing them a second time is a lot less enjoyable. I took the time to run the network cables from my switch through to most rooms in the house under the floor, and used proper terminators and wall blocks at each end. Compared to the alternative (stringing cables through the house, fastened down with duct tape) it’s a lot better to look at, and less things can go wrong. Apply the same to all areas of networking.
Patch panels and the like

If you’ve already got a rack (or are now thinking about getting one), think also about getting a patch panel. It allows you to bring all the cables from areas of the house into one area of your rack, which you can label and keep track of.

Conclusion

Those are pretty much all the things I’ve kept in my head as I built my network. I’d love to hear more of your suggestions on what to do and consider as your network goes. If you’d care to leave abuse for me suggesting the use of an expensive rackmount setup, that’s also fine (but one day, you will think the same).

[ad#post-footer]

I’m currently in the process of building myself a new home file server, to act as a storage medium for my MP3 music collection, my father’s business document backups, as well as to do compiles of software projects I’m working on.

For this, I’ve basically taken my old desktop computer (sans high-end graphics card) and place it in an Antec rack-mount case. After adding an expensive RAID card and a few hard drives, I have the hardware ready to go (though I am still contemplating the purchase of a PCI-based ADSL interface card).

Naturally, Debian was the first choice of operating system. Although I toyed with the idea of Windows Vista Business enough to almost go out and buy a copy, I decided:

1. It wasn’t worth the $140 just to have the use of Remote Desktop instead of SSH.
2. In the future, I may also want to use the server as an ADSL gateway, wireless access point, compile box, TeX environment, web server, etc. Vista can’t be made to do any of that easily.

Installing Debian was an interesting experience. I originally had the two 80GB Seagate hard drives installed on the onboard JMicron RAID controller (to run in RAID 1 for extra dependability). I couldn’t find drivers for this chipset that my Debian install disc would like. I then changed to using the JMicron controller in plain IDE-compatibility mode. Debian picked up only one hard drive. I’m still bemused by how this is possible: both the BIOS and a UBCD4Win disc pick up two. I then switched the hard drives and plugged them into the ICH9 controller. Debian picked up both of them, and so far is working fine. Although the ICH9 chip on my motherboard (a GA-P35-DS3) doesn’t support RAID, I’m still fairly happy. Anything I really want to keep will be stored on the RAID set managed by the RocketRAID card.

During the install process (and with all these changes of plan), it was great to have one of my Linux-savvy friends, Chris Neugebauer, on hand. Unfortunately, the same cannot be said of the #debian channel on Freenode. On both the occasions I asked questions, I was just plain ignored. Now, I know they don’t owe me anything, but a simple “Sorry, no idea, you’re on your own” would have been nice.

But apart from that minor downside, both the hardware and software are going really well so far. I’ll probably follow up with another article once I’ve finished further my home network. I still want to buy a few more parts…

I stumbled across this cool thing on the Internet the other day, called the World Community Grid (WCG). It is basically the next step forward from SETI@home. For those of you who don’t know, SETI@home was (and still is) a distributed network application that calculated the probabilities of aliens existing, from readings from radio telescopes and the like. It did this in the idle time those node computers had. The thought to use the combined power of hundreds of thousands of computers was a brilliant idea. The idea to waste that combined power on searching for aliens (who MUST exist, due to sheer probabilities and the size of the universe) was not.

(It is also a fact that due to the sheer size of the universe, and the time said universe has been in existence, it is impossible for those aliens to visit us on Earth. Unfortunately, the crop circle people have not yet realised this.)

Anyway. We have this brilliant idea to combine the computing power of lots of idle machines. So somebody says, “cure AIDS!”, and somebody else says “cure Cancer!”, and then somebody else finally says “cure African climate problems!”. So the WCG was set up to manage the distribution of work between all the participating machines, on behalf of the projects that use it.

Of course, the question must be raised whether the benefit gained by leaving your computer on, idle, is greater than the benefit gained by turning your computer off (or into low-power mode) when it is not in use. It’s a hard decision, and I guess it depends on what kind of power supply you have, how much you pay for it, and how often you can bear to wait for your computer to start up. For me, the answers are: hydro power, not that much (relative in global terms), and not very often at all. So I leave my computer on all the time (which is great for my WCG rank), and just turn the monitors off (if you have two CRTs like I do in my bedroom, this alone could save the world).

Needless to say, I joined up. I hate wasting machine cycles. I currently have two computers connected under my name, and I am hoping to add a couple more soon. I would love all the people reading this to also at least consider signing up. If you do, leave a comment with your username, so I can have a look at your statistics every so often.