Web Servers

1. Confirm That It's a Bandwidth Problem

As was stated above, not every report of a slow network connection is due to a bandwidth issue, but just about every bandwidth issue results in slow network performance for at least some users. So, if the initial report is slow network performance, we must first determine if it's related to bandwidth oversubscription or not. Fortunately, with the wide range of choices out there for network monitoring solutions, some based on flow-based technologies (NetFlow, sFlow, etc.) and some that are packet-based, correlating a problem report with overall network activity has become fairly straightforward. One simply needs to "rewind" the network activity to the period of time where slow network performance was reported, isolate the flow where the event was experienced, and then expand the view to include data collected at the router or switch level for the flow to see if the network link was oversubscribed. If so, a bandwidth issue of some sort is probably the cause, and a more detailed investigation is in order. Alternatively, most network monitoring solutions provide flexible alarms and notifications based on measurements like percentage of bandwidth utilization on a link. If a triggered alarm corresponds to the time of the report of poor performance, you've already answered the first question.

2. Determine If It's a One-Time Event

If your network monitoring solution indicates that a bandwidth threshold has been exceeded, the next thing to determine is the frequency of such an event. Is this the first time, or have similar incidents occurred in the recent past? If it is the first occurrence, and especially if it is still ongoing, you should dig in quickly to see who the top talker is. Though the event could be benign, like too much video being downloaded by too many users on the day of a popular event, it could also be an indication of something serious, like a distributed denial of service attack or a worm infection. Regardless of the cause, a spike in bandwidth that is unpredicted and outside of the norms for your network should be taken seriously and addressed immediately.

If similar bandwidth spikes have occurred before, or if you've already determined that certain spikes are part of the norm (eg. order processing spikes first thing in the morning when the operations department starts their day), make sure that the current spike conforms to one of these norms. The best way to determine this is to have current baseline measurements of typical network usage, since it is highly unlikely that network usage is constant throughout the day. Understanding what the load is at different times of the day, and what the peak historical usage has ever been, will help you tremendously in determining the severity of the next spike in bandwidth usage.

3. Remove Unnecessary Traffic

More often than not, bandwidth problems come from unnecessary traffic on the network, like real-time video streaming during the 7th game of the World Series. By pruning your WLAN traffic and removing or reducing superfluous demands through the use of bandwidth shaping technologies you can better control overall bandwidth usage and ensure that adequate bandwidth is always available for your mission critical applications.

These three steps should help you determine when and where bandwidth issues occur and how you can better manage your network to stop slow network performance caused by network bandwidth spikes.

*help

This command gives you a key to virtually everything you need to use DOS effectively. Can't remember the format for the FORMAT command? Not a problem! Type HELP FORMAT and press the ENTER key, and you've got everything you need to use FORMAT! This is true for virtually every DOS command so if you remember HELP command name you won't have any problems.

Another way to get help (or, if you don't install the HELP files suggest above, the ONLY way to get help in Windows 95) is to use /? (slash, question mark) after any command. This is faster and sometimes a more useful form of help, but it is generally only a listing of options. HELP is usually a more verbose help facility than /? is.

*copy

Once you know where to get help, the rest is easy. The whole purpose of DOS is to let you get around the directory tree easily. You use DOS to start applications (like my own P11.EXE program) and to copy files, delete files, build places to put files, and compare files, and so on. Files, Files, Files. That's what DOS is really for.

The #1 tool for file handling is the COPY command. Specify the "from" file first, then the "to" file. COPY only copies from within one directory at a time, but the "to" directory can be different. For copying entire directory structures, use XCOPY instead.

What makes COPY so powerful is that you can use "wild cards" like * and ? within the file name.

*xcopy

Like COPY, XCOPY can use wildcards to copy a file or group of files to another directory. But XCOPY is more powerful because it has the ability to copy an entire directory structure in one command. Use the /s option to accomplish this. Use HELP XCOPY to learn more, of course!

*cd

CD is the cold, hard way of doing things, but sometimes it's the right way. What makes it cold and hard is that in comparison to Norton's Change Directory (normally abbreviated NCD, but I abbreviated it to C on my system), the DOS CD command requires you to type the complete directory entry, all the way back to the root level! This is absurd. NCD, on the other hand, lets you type in anything that's close, and it will search your entire directory structure and plop you down right where you want to be. For example, to go to the \p11\pumps\svgapump directory, I can just type in: c svgap and immediately NCD will take me there. With CD, I would have to type cd \p11\pumps\svgapump to accomplish the same thing. Saving keystrokes is the name of the game. (That's why later in this tutorial, you'll learn to write batch files.)