NETWORK MALFUNCTIONS

Any malfunction of the network is going to result in a nonavailability of the system to the users. The diagnosis and fixing of this malfunction becomes a high priority. There are three primary culprits to network malfunctions: component and server failures, and data collisions.

COMPONENT FAILURE

Component failures are categorized in two categories: hard faults and soft faults. Hard faults are relatively easy to find, and a diagnostic program will diagnose them correctly every time. Soft faults can be difficult to find, because they occur sporadically or only under specific circumstances, rather than every time the memory location is tested. A diagnostic program tests computer hardware and peripheral devices for correct operation.

Most computers run a simple set of system checks when the computer is turned on. The PC tests are stored in read-only memory (ROM), and are known as power-on self tests (POSTs). If a POST detects an error condition, the computer will stop and display an error on the screen. Some computers will emit a beep signal to indicate the type of error.

One of the best tools to use for network malfunctions is a network analyzer. A network analyzer is a product that can be used to monitor the activity of a network and the stations on it, and to provide daily summaries or long-term trends of network usage and performance. A network analyzer can do tasks such as:

l Count or filter network traffic.

l Analyze network activity involving specified protocols or frame structures.

l Generate, display, and print statistics about network activity, either as they are being generated or in summary form.

l Send alarms to a network supervisor or network management program if any of the statistics being monitored exceeds predetermined limits.

l Do trend or pattern analyses of network activity.

Network analyzers may be software only or consist of both software and hardware. The latter may include an interface card enabling you to test the network directly. This card may include an on-board processor. Because of their greater capabilities, hardware/software analyzers are more expensive than the software only analyzers. In fact, the prices for the hardware/software analyzers can be several times as high as those for the software only versions.

SERVER FAILURE

The most obvious sign that the server has failed for some reason is that all users, except root, will not be able to logon to the system. Use the following steps as required to reestablish services:

The first and easiest thing to try is to run the system distribution again. This will rebuild the system maps if nothing else is wrong and will allow users access to the system.

Shutdown and reboot the system. During the boot process ensure that no failures occur on any of the nodes.

l Verify the domain name.

l Look for the maps subdirectory; it should be the same as the domain name. If it is not there, you will need to run the system initialization command.

l If the above fails, ensure that all the files to be mapped are present on the server. If any have been deleted, they will have to be restored from the latest system saves.

One of the best ways to avoid server malfunctions is to conduct maintenance on the server. It is important to set up a schedule for your server and strictly adhere to it. To check the hardware, you should do at least the following things:

l Clean the server carefully but thoroughly.

l Check cabling and connections for tightness and signs of bending or stress. Do not disconnect connectors unless necessary, since many connectors are rated for a limited number of matings.

l If possible, check the cabling with a line analyzer.

l Run thorough diagnostics on the storage medium and on other system components to identify the components that are likely to fail and to deal with these before they actually do fail.

l Check the quality of your power line by using a line tester.

The hardest part of server maintenance is finding the time to conduct the maintenance, since the network will have to be offline to conduct. In many cases, server maintenance will need to be during off peak hours, late night or early morning, when there is little or no use.