Research Of Network Neutrality Based On Network Tomography And Packet Loss Rate

Network neutrality is a rule made by the government. It stresses that Internet Service Providers (ISPs) should equally treat each legal user. And they are forbidden to block or throttle any user's legal behavior, even set higher priorities. The aim of network neutrality is to build a fair and healthy network.We analyze network neutrality from parts to the whole by quantity. Network Tomography is used to infer properties of links by external observations. In this paper, we design an improved algorithm based on maximal flow algorithm to judge whether a link sequence is neutral or not using Network Tomography. We filter out link sequences which are beyond link capacities. This improves True Positive Rate (TPR) and reduces False Negative Rate (FNR). The improved algorithm has no false cases. Through simulation experiments, we find that when there is only one non-neutral link, TPR of the original and improved algorithm both first increases, then decreases while the number of the paths increases. The improved algorithm has an average increase of about 57%, up to 61.66% of TPR compared with the original. When there is 2 or 3 non-neutral links, TPR of the original algorithm is relatively stable, and TPR of the improved algorithm improves while the number of paths increases. The improved algorithm has an average increase of about 8% of TPR compared with the original.Based on the partial analysis of network neutrality, we analyze neutrality of the whole network by external observations. Packet loss rate (PLR) is an important metric to evaluate network condition. Based on the model for calculating PLR, we use data provided by Neubot and the proposed elementary function model, and come to a conclusion that the threshold of PLR is 0.10. The difference of PLR of two tests is used to evaluate the level of network neutrality violations. Based on the threshold of PLR, we analyze the level of neutrality violations in time and space. In time, neutrality violations are more likely to happen in peak hours. In space, the level of neutrality violations is highest in China and South Africa among the eight regions studied. Finally, we analyze PLR of 2012 to 2015 worldwide, and find the relationship between the speed of network infrastructure development and user demand.