Useful Traffic Loss Rate Estimation Based on Network Layer Measurement

The current network is information asymmetric, about what happens in the network, and end users know much less than the Internet service provider. Especially, in a non-neutral network, for commercial interests, the operator's behavior of adjusting network configuration and controlling data transmission priority will seriously infringe the vital interests of end users. In order to rebalance such information asymmetry to empower users, researchers have been committed to the study of end-user network performance measurements. In this paper, we contribute to the ongoing efforts to empower users by defining and estimating the useful traffic loss rate (UTLR) experienced by a Transfer Control Protocol connection. The work presented in this paper is inspired by our empirical observation that what the operator's behavior ultimately affects is the link quality, and therefore, a good characterization of the impact of link quality changes on user's quality of experience (QoE) is the prerequisite for empowering users. To this end, the metric UTLR is first defined in this paper to better characterize the impact of link quality changes on user's QoE, and then, we construct a series of heuristic rules with the receiver network layer information to estimate UTLR. To improve the estimation accuracy, the effects of insufficient duplicate ACKs, static round trip time, enlarged loss period, repeated packet losses, and identification field cyclic reuse are also considered. Finally, the gradual and rigorous experiments show that the proposed method achieves a stable and precise estimation for UTLR over the entire measured parameter space.