Network Performance Issues

One of the basic requirements of a computer network is connectivity. Without a connected network, the computers may not communicate with one another. This requirement leads to different network topologies and interconnections. Another feature of a network is its capability to share resources.

A network allows the sharing of disk space, printers, and other hardware and software. In addition to being cost effective, resource sharing allows the incremental growth of the network while providing leverage for the current investment.

The network designer can effectively increase the network size without discarding the currently utilized resources. The study of network performance allows a better understanding of network connectivity and resource sharing.

For instance, a comparison may be made between the delays to run an application on a workstation versus running it on a server, which may lead to a decision of buying the required number of application licenses. Similarly, one LAN topology may be preferred over another, based upon performance.

Currently, businesses and organizations debate the advantages and flexibility of deploying a wireless network versus a wired network. A requirements analysis that addresses the performance issues may help in taking this decision.

Performance of computer networks may be evaluated using three main techniques: analytical modeling, simulation, and direct measurements. Because these techniques complement each other and may have their own limitations, often their combination is used to predict the system performance under certain conditions.

Often, the r esults from one approach confirm the results obtained from another if the models are built correctly. On the other hand, the problematic models may be fixed, the error often arising due to an incorrect assumption, incorrect use of a variable, or incorrect flow of data/control in the model.

Network mathematical analysis is often performed for a specific scenario with limited and very rigid conditions. Although simulation models usually provide a better understanding for design and development of networks, it requires software expertise with a deep understanding of network operation.

Direct measurement allows thorough network troubleshooting, but network expansion and planning may be limited. Network emulation refers to the ability to introduce the simulator into a live network.

Special objects within the simulator are capable of introducing live traffic into the simulator and injecting traffic from the simulator into the live network. In this approach, live traffic may be passed through the simulator and the effect on it noted by objects within the simulation or by other traffic on the live network.

In another approach, the simulator can include traffic sources or sinks that communicate with real-world entities. Using this combination of simulation and direct measurements is often very helpful in making performance predictions under “what-if” scenarios.