Analysis of the YouTube server selection behavior observed in a large German ISP network

Abstract

This paper presents a detailed study on the YouTube server selection mechanism with respect to its long term behavior. We conducted a measurement campaign from different measurement points within a single German ISP network over a long period of time. Based on the measurement traces, the longterm temporal behavior of the server selection mechanism is investigated. The results reveal that the server selection behavior ranges from changing on a daily basis to being stable for several weeks. Besides analyzing the long term behavior, we also determined the dependency of the selection behavior on the geographic location of the client node. It appears that the server selection behavior mainly depends on the IP address currently assigned to a user connection. Finally the impact of the selection decision on the round trip time (RTT) is analyzed.