Design and implementation of a network traffic analyzing utility

Abstract

Developments in Information and Communication Technology (lCT) have created many new applications, which require a large amount of Internet bandwidth for proper application usage. However, Internet bandwidth is an expensive resource especially in this part of the world. Therefore, enterprises that use the Internet for business always need to efficiently use the bandwidth. Internet Service Providers may also need to manage their Internet bandwidth efficiently in order to increase the profit margin of their services. Knowing the current (and past) usage is an important requirement in estimating the bandwidth requirement of a corporate network. The information about the composition of the use of the bandwidth by applications, user groups, etc. will help an administrator to efficiently manage it. Furthermore, knowing what is happening in the network always helps an administrator to keep the network secure, efficient and reliable all the time. We studied the requirements of a network administrator by considering LEARN as a test bed for this study. We then assessed, whether the existing network traffic monitoring tools can provide the required information to the administrator. We examined some of the popular network monitoring tools that are widely used in the Internet community for their features and drawbacks. We found that the existing tools for network traffic monitoring are not capable of providing most of the required information by the administrator. With these results, we identified the features available in those tools, and developed a new tool, LEARNStat, re-using some of the freeware utilities available in order to meet the requirements of a network administrator. We tested LEARNStat in LEARN which is our test-bed, and during the short period of running LEARNStat, we were able to obtain several important results. In this thesis, we discuss the requirements of a network administrator and how we met those with LEARNStat. We present a brief description of related principles and also discuss the results we obtained with LEARNStat. The thesis also includes possible future enhancements for the LEARNStat.