Accurately characterizing end-to-end Internet dynamics - the performance that a user actually obtains from the lengthy series of network links that comprise a path through the Internet - is exceptionally difficult, due to the network`s immense heterogeneity. At the heart of this work is a `measurement framework` in which a number of sites around the Internet host a specialized measurement service. By coordinating `probes` between pairs of these sites one can measure end-to-end behavior along O(N 2) paths for a framework consisting of N sites. Consequently, one obtains a superlinear scaling that allows measuring a rich cross-section of Internet behavior without requiring huge numbers of observation points. 37 sites participated in this study, allowing the author to measure more than 1,000 distinct Internet paths. The first part of this work looks at the behavior of end-to-end routing: the series of routers over which a connection`s packets travel. Based on 40,000 measurements made using this framework, the author analyzes: routing `pathologies` such as loops, outages, and flutter; the stability of routes over time; and the symmetry of routing along the two directions of an end-to-end path. The author finds that pathologies increased significantly over the course of 1995 and that Internetmore » paths are heavily dominated by a single route. The second part of this work studies end-to-end Internet packet dynamics. The author analyzes 20,000 TCP transfers of 100 Kbyte each to investigate the performance of both the TCP endpoints and the Internet paths. The measurements used for this part of the study are much richer than those for the first part, but require a great degree of attention to issues of calibration, which are addressed by applying self-consistency checks to the measurements whenever possible. The author finds that packet filters are capable of a wide range of measurement errors, some of which, if undetected, can significantly taint subsequent analysis. « less