Abstract The 2017 FCC Restoring Internet Freedom Order removes the “enhanced” transparency obligations introduced by the 2015 Open Internet Order and aims to return net neutrality policy to transparency rules based on the 2010 Open Internet Order. The ruling states that the burden of additional network performance disclosures exceed the benefits, and that the most salient metrics to report are those that involve consumer quality of experience for the applications that they commonly use. Unfortunately, however, Internet service providers (ISPs) will typically have difficulty reporting on application performance and QoE metrics, both of which are notoriously difficult to estimate from network traffic.
To address this shortcoming, we introduce a new tool that estimates QoE for Internet video streaming from passively collected network traffic. The tool we develop sits inline, on path, and analyzes traffic in real time as it traverses the network to (1) identify which traffic flows belong to a specific video streaming service; (2) estimate four critical quality-of-experience metrics for streaming video: bitrate, changes in bitrate, join time, and rebuffering. When deployed on a small embedded device, the tool operates at traffic of near 1 Gbps, making it suitable for deployment in consumer home networks, as well as near various network endpoints.
Because Internet video traffic accounts for majority of the global internet traffic, this approach of passively observing traffic has two significant policy implications: 1. It reduces the administrative and operational burden on ISPs, because traffic collection and analysis is passive, in-line, and in homes, and does not introduce additional test traffic. 2. The approach offers application QoE metrics that are complementary to the lower-level network performance metrics that ISPs already collect.
We discuss the capabilities of the tool, its initial deployment to a large number of consumer homes, our initial findings concerning the reporting of application QoE metrics, and broader implications for policy surrounding ISP transparency reporting requirements. The tool we have developed can provide a deeper understanding of several concepts often discussed both in the context of net neutrality and incentivizing competitive forces in the market.
First, this tool can shed more light on the nature of streaming traffic from consumers, including which video streaming services are most popular with consumers, and how those popular services perform on different networks. Measurements based on service-specific usage is particularly meaningful to consumers because consumers often understand their network needs better when it is tied to the applications they use often. More specific information about the performance of popular video services can help consumers make more informed choices about the network services that they purchase. Second, the tool facilitates the analysis of application performance for network traffic at multiple locations along a single end-to-end path, enabling both consumers and regulators to independently verify ISP reports about application performance. Ultimately, the type of information about application performance that our tool exposes can affect consumer decision-making; we explore and discuss these effects, and how they may ultimately interact with switching costs, market competition, and other commercial considerations.
