The problems with Internet measurement
The Internet is a wide-ranging collection of networks, which as a whole has chaotic behaviour , characteristics. In order to understand, model and predict this behaviour, some type of accurate measurement is needed. One form of measurement is active measurement; this involves introducing packets into the network and measuring the way the network handles those packets. This paper focuses on packet delay measurements. Current techniques for measuring delay have certain issues related to their reliability. The objective of this report is to highlight the issues associated with packet probing measures and to propose a solution that in many cases improve the ability to measure more accurately where and when problems occur within the modern Internet.
The structure of this report will look into the difficulties associated with host clock synchronisation and how it affects some measurement protocols. Furthermore current packet probing techniques such as ping and traceroute will be examined, looking at both the advantages and problems associated with using these measures.
Continuing on from the current inefficiencies of measuring techniques, a network measuring protocol will be proposed as a solution to the problems found.
Most packet probing techniques that measure delay to sections of a network, such as one-way delay measurement require synchronized clocks at every host on the route (Shalunov, 2001). The fundamentals of clock theory are that clocks are of limited precision and that they drift at differing rates over time (Comer, 2000), Although there are several ways accurate timestamps can be generated. One way is to provide an accurate, machine-readable clock synchronized to Global Positioning system (GPS) at every service point with the Code Division Multiple Access (CDMA) network. An accurate measurement obtained with GPS synchronization provides optimal results, however its implementation it is not so simple...