| Previous | Table of Contents | Next |
“We only truly learn by destroying.”—Usenet post, comp.sys.unix (circa 1980)
Network geeks don’t live in a vacuum; we all have to learn how to deal with the things our networks live on. Whether it’s a bad circuit board or a corrupt spot on a hard drive, network functions are picky, picky, picky; they stop working when underlying pieces and parts (the infrastructure) stop working. Networkers have been fixing (and accidentally breaking) hardware for years. (People may think all we do is stare at scores of monitors all over our network command center and stroke our beards thoughtfully, but we’ve really got our sleeves rolled up. We’re the equivalent of a silicon grease monkey.) Once a problem has been troubleshot down to a hardware component, hardware troubleshooting is often necessary. In this hour, I’ll cover what to do and what not to do with hardware. I’ll also go over the basics of the following topics:
Because many servers are nothing more than pumped up PCs with faster and better hard drives, CPUs, and so on, many of these hardware techniques apply to servers as well.
A user’s PC is often the culprit of a network problem. Although a PC is essentially just an assemblage of pluggable circuit boards and chips and can be troubleshot from that angle, you should also consider its low-level software configuration when troubleshooting. Your PC’s lowest level software is its BIOS (Basic Input/Output System), which is responsible for making all the boards and chips talk to each other in a civil manner.
You can get into a PC’s BIOS by pressing the Del, F1, or F2 key right after you power on the PC. (Ctrl+Alt+Esc used to be popular, but I haven’t seen this one in quite some time.) The BIOS setup screen can range from fairly complex to reasonably simple. Typically, a name-brand PC’s BIOS is simpler than a generic clone’s, because the name-brand manufacturer exercises greater standardization over which components are connected to its motherboards. A typical BIOS screen is text-based, with no Windows-style controls.
A BIOS setup screen can be compared with other “known good” PCs from the same manufacturer in order to verify that the settings are correct. If you don’t have another PC of the same type, you can always reset to the defaults. This can often correct a problem that some nut behind the keyboard might have caused.
For example, certain memory settings are configurable at the BIOS screen, as are certain Plug and Play settings. If one of these is changed to an incorrect value, your system may start malfunctioning in the most interesting of ways. In particular, certain non-Plug and Play network cards, for example, have limited IRQ (Interrupt Request) numbers they can use; if the BIOS reserves these IRQs for Plug and Play devices, your network card will either not work at all or will behave erratically.
Be sure to write down your BIOS settings before you reset to the defaults, because you might need some of those settings later. Typically, resetting your BIOS to its defaults won’t hurt anything, but you never can tell.Some PCs have the option to print BIOS settings to a local printer—if yours does, go for it. Printing out your settings beats the heck out of writing everything down.
The components within most PCs are reasonably consistent:
The differentiating factor between PCs tends to relate to the physical layout of components. My experience has shown that cheap clones, as opposed to name-brand machines, can be poorly designed, which can become a factor in situations requiring you to troubleshoot. Suppose, for example, that the PC you’re troubleshooting sports a poorly placed motherboard jumper block that comes into contact with an expansion card. (A jumper is a movable and removable mechanism that electrically ties two pins together on the motherboard; a jumper block is a group of these pins, usually controlling some kind of configuration of the motherboard.) All of a sudden, a jumper set that wasn’t connected becomes connected through the metal of the expansion card bracket, and chaos ensues.
Clone PCs are often not compatibility tested with networks the way name-brand machines are, which can add to the fun. The bottom line is this: Spending a few extra dollars for a name brand can help you avoid problems.Of course, buying a brand name doesn’t always mean that things work 100 percent of the time, and you still might run into network issues. However, it’s important in a networked environment with large numbers of PCs to be able to have some sort of accountability when things don’t work as promised—which you don’t often get when buying from the here-today-gone-tomorrow clone market.
| Previous | Table of Contents | Next |