Master's Thesis of Abhishek Kapoor
We present an active measurement test (netflix) that down- loads and mimics playout of content from the Netflix content delivery network. The test measures TCP connect times, prebuffering duration, achievable throughput, number and duration of stalls as indicators of performance when down- loading content from Netflix. We deployed netflix on ∼100 SamKnows probes connected to dual-stacked networks representing 66 different origin ASes. Using a 22-months long dataset (July 2016 - May 2018) we observe that besides some vantage points that experience low success rates to connect over IPv6, Netflix Open Connect Appliance (OCA) infrastructure appears to be highly available. We observe that clients prefer connecting to Netflix OCA over IPv6, while the preference tends to drop over certain peak hours during the day. The latency towards the OCA has reduced over the year with the TCP connect times being comparable over both address families, but it takes longer to buffer the con- tent over IPv6. Achievable throughput has increased over the year, although higher number of stalls over IPv6 tends to reduce the bitrate and consequently the achievable through- put over IPv6. The durations of the stalls are lower over IPv6. We also provision scamper right after the netflix test to capture the forwarding path towards the Netflix OCA. We observe that the Netflix OCA deployed inside the ISP are reachable within 5 IP hops. We observe that such ISP caches can reduce TCP connect times by ∼10 ms and pre- buffering durations by ∼500 ms over both address families. The achieved throughput can also increase by a factor of four when such ISP caches are in use. We observe that more number of paths towards Netflix OCA are shorter over IPv6. Shorter path lengths do not always correlate with lower la- tency as seen with ISP caches that are only available over IPv6, where at times lower latency is still experienced to- wards the OCA deployed inside the Netflix Content Delivery Network (CDN) over IPv4. This is the first study to measure Netflix from residential networks, since the inception of the Netflix CDN infrastructure in 2011.