Hits:

Indexed by:会议论文

Date of Publication:2015-11-04

Included Journals:EI、CPCI-S

Page Number:297-304

Abstract:Network neutrality is a policy made by the government. It emphasizes that Internet Service Providers (ISPs) should treat all users equally, regardless of users' behavior, as long as they are legal. In this sense, it aims to build a fair and healthy network. Packet Loss Rate (PLR) is an important metric to evaluate a network's condition. In this paper, we compare PLR of two different kinds of tests, which are "speedtest" and "bittorrent", to judge network neutrality of ISPs. The difference of PLR of two tests is used to evaluate the level of network neutrality violations. We propose a general framework for analysis of network neutrality. First, by transforming the well-known Mathis model, we build a model that uses public data, collected by the network neutrality bot (Neubot), to estimate the PLR of countries all over the world. Second, a composite function model, named PECF, is proposed for fitting to get the lower limit of PLR (L-PLR). When the value of PLR is LPLR, the frequencies of PLR of two tests differ most and a network neutrality violation appears. By mathematical inference, we draw the conclusion that ISPs violate network neutrality when PLR is larger than 0.10. Third, based on this conclusion, we concentrate on the case when PLR is larger than 0.10 to study network neutrality violations. We use R-PLR (Ranged PLR) to describe PLR within a certain range. By comparing the PLR of "speedtest" with that of "bittorrent" within a certain R-PLR, we further determine the level of network neutrality violation in time and space. Network neutrality violation is more likely to happen under network peaks. The geographical distribution of PLR shows China (CN) and South Africa (ZA) suffer the most severe network neutrality violations among the eight regions studied. Finally, we present the histogram of PLR worldwide from year 2012 to 2015, which shows a relationship between network development and user demand.