Estimating packet loss rate in the access through application-level measurements

Abstract

End user monitoring of quality of experience is one of the necessary steps to achieve an effective and winning control over network neutrality. The involvement of the end user, however, requires the development of light and user-friendly tools that can be easily run at the application level with limited effort and network resources usage. In this paper, we propose a simple model to estimate packet loss rate perceived by a connection, by round trip time and TCP goodput samples collected at the application level. The model is derived from the well-known Mathis equation, which predicts the bandwidth of a steady-state TCP connection under random losses and delayed ACKs and it is evaluated in a testbed environment under a wide range of different conditions. Experiments are also run on real access networks. We plan to use the model to analyze the results collected by the "network neutrality bot" (Neubot), a research tool that performs application-level network-performance measurements. However, the methodology is easily portable and can be interesting for basically any user application that performs large downloads or uploads and requires to estimate access network quality and its variations.