If you have been following my articles, you probably have a fair idea of what network congestion is and its probable causes. In brief, we said that "network congestion” is a state of user-visible degradation of network performance.

It occurs when network assets are not able to process data as fast as users may wish, which makes this definition rather relative. We also said that Network Congestion is caused by either overload on the individual network assets or limited capability of the assets or both.

Depending on what a network is used for, network delays may or may not be categorised as congestion and this is where the ICT wizards may take us for a ride. As long as the network is maintained within the congestion-prone level or below, one may argue that the network does not warrant instituting congestion management solutions, some of which may be expensive.

Network Managers that are not customer focused will only take measures when network jamming occurs; that is, when data loss is experienced.

A case in point is the internet café. A network of twenty high-speed computers connected to a very low speed switch and low speed interface will cause network congestion while a similar network with only ten computers will not.

Do not be fooled by the beautiful furniture! On the other hand, if the ten computers are all used to stream music and video, they could lead to congestion compared to the twenty ,if used for IM or email applications only.

The same thing applies to your organization network. If all users decided to download heavy files at the same time, network congestion is likely to occur irrespective of how modern your network is.

Fortunately most modern network assets have in-built mechanisms that may enable the network to avoid jamming.

Solutions to network congestion are based on either avoidance or control principles; to use the layman’s language, the preventive and curative principles respectively.

Whereas the control solutions may be easy to conceive since one is dealing with known parameters, avoidance becomes challenging mainly because the cause of congestion will not be known.

Congestion avoidance solutions are those that will enable the network to operate below the congestion-prone region while congestion control ones are those that will provide a recovery mechanism once a network has entered into a congestion mode.

Although switching assets in a network have inbuilt mechanisms for individual asset congestion management, end-to-end congestion management remains a challenge mainly because of its dynamic and distributed nature.

Existing engineering solutions are based on dynamically increasing the available network resources or dynamically decreasing the demand on available resources.
 
New technologies, however, have enabled network users to get access to more data and to demand for more speed, reliable and fair delivery. This puts a challenge to engineering solutions.

Increasing available resources becomes limited by both cost and available technologies while decreasing the demand amounts to service degradation, including unfair distribution of resources.

When users put varying data loads on the network, congestion becomes unpredictable, which puts more challenge to engineering solutions. The unpredictable loads will result into some assets being underutilised when the demand is low or available assets being overloaded when the demand is high.

In the next issue we shall look at the more cost effective solutions, based on users.

The writer is the Managing Director of Cornerstone Africa Ltd
www.cornerstone-africa.com
info@cornerstone-africa.com