We present an active measurement test (netflix) that downloads content from the Netflix content delivery network. The test measures latency and achievable throughput as key performance indicators when downloading the content from Netflix. We deployed the test on ~100 SamKnows probes connected to dual-stacked networks representing 74 different origin ASes. Using a ~2.75 year-long (Jul 2016-Apr 2019) dataset, we observe Netflix Open Connect Appliance (OCA) infrastructure to be highly available, although some vantage points experience low success rates connecting over IPv6. We witness that clients prefer connecting to Netflix OCAs over IPv6, although the preference over IPv6 tends to drop over certain peak hours during the day. The TCP connect times toward the OCAs have reduced by ~40% and the achievable throughput has increased by 20% over the measurement duration. We also provision scamper right after the netflix test to capture the forwarding path toward the Netflix OCAs. We observe that the Netflix OCA caches deployed inside the ISP are reachable within six IP hops and can reduce IP path lengths by 40% over IPv4 and by half over IPv6. Consequently, TCP connect times are reduced by 64% over both address families. The achieved throughput can~ also increase by a factor of three when such ISP caches are used to stream content. This is the first study to measure Netflix content delivery from residential networks, since the inception of the Netflix CDN infrastructure in 2011. To encourage reproducibility of our work, an anonymized version of the entire longitudinal dataset is publicly released.

We compare the similarity of webpages delivered over IPv4 and IPv6. Using the SamKnows web performance (webget) test, we implemented an extension (simweb) that allows us to measure the similarity of webpages. The simweb test measures against ALEXA top 100 dual-stacked websites from 80 SamKnows probes connected to dual-stacked networks representing 58 different ASes. Using a two months-long dataset we show that 14% of these dual-stacked websites exhibit a dissimilarity in the number of fetched webpage elements, with 94% of them exhibiting a dissimilarity in their size. We show that 6% of these websites announce AAAA entries in the DNS but no content is delivered over IPv6 when an HTTP request is made. We also noticed several cases where not all webpage elements (such as images, javascript and CSS) of a dual-stacked website are available over IPv6. We show that 27% of the dual-stacked websites have some fraction of webpage elements that fail over IPv6, with 9% of the websites having more than 50% webpage elements that fail over IPv6. We perform a causality analysis and also identify sources for these failing elements. We show that 12% of these websites have more than 50% webpage elements that belong to the same origin source and fail over IPv6. Failure rates are largely affected by DNS resolution error on images, javascript and CSS content delivered from both same-origin and cross-origin sources. These failures tend to cripple experience for users behind an IPv6-only network and a quantification of failure cases may help improve IPv6 adoption on the Internet.

We present the first large-scale analysis of Web performance bottlenecks as measured from broadband access networks, using data collected from extensive home router deployments. We analyze the limits of throughput on improving Web performance and identify the contribution of critical factors such as DNS lookups and TCP connection establishment to Web page load times. We find that, as broadband speeds continue to increase, other factors such as TCP connection setup time, server response time, and network latency are often dominant performance bottlenecks. Thus, realizing a "faster Web" requires not only higher download throughput, but also optimizations to reduce both client and server-side latency.

We present the first study of network access link performance measured directly from home gateway devices. Policymakers, ISPs, and users are increasingly interested in studying the performance of Internet access links. Because of many confounding factors in a home network or on end hosts, however, thoroughly understanding access network performance requires deploying measurement infrastructure in users' homes as gateway devices. In conjunction with the Federal Communication Commission's study of broadband Internet access in the United States, we study the throughput and latency of network access links using longitudinal measurements from nearly 4,000 gateway devices across 8 ISPs from a deployment of over 4,200 devices. We study the performance users achieve and how various factors ranging from the user's choice of modem to the ISP's traffic shaping policies can affect performance. Our study yields many important findings about the characteristics of existing access networks. Our findings also provide insights into the ways that access network performance should be measured and presented to users, which can help inform ongoing broader efforts to benchmark the performance of access networks.