Measuring the Efgects of Happy Eyeballs Takeway Limitations - PowerPoint PPT Presentation

Introduction Motivation Leone Project: leone-project.eu Supported by: October 2016 Jacobs University, Bremen Jürgen Schönwälder Joint work with Applied Networking Research Workshop Published at: Madrid, Spain RIPE 73 Jacobs University, Bremen Vaibhav Bajpai Measuring the Efgects of Happy Eyeballs Takeway Limitations Measurement Setup Browser Implementations Research Contributions Methodology Metrics and Implementation Selection of Websites 1 / 18 Lowering HE Timer Measurement Trial Data Analysis Trends Who connects faster? Preference Slowness July 2016: https://goo.gl/ZiIcQV Flamingo Project: fmamingo-project.eu

Introduction Limitations 2% 3.5% 2015 5.3% 40% 2011 Native Overall 2015 Tie 6to4 anycast prefjx has been obsoleted [6]. 2013 Microsofu stopped Teredo on Windows and deactivated public Teredo servers [5]. Motivation Takeway Introduction | Motivation Lowering HE Timer Data Analysis Browser Implementations Research Contributions Methodology Metrics and Implementation Selection of Websites Measurement Setup Measurement Trial 2 / 18 Slowness Trends Who connects faster? Preference ▶ HE timer (300 ms) was chosen (2012) when broken IPv6 connectivity was prevalent. ▶ Largely attributed to failures caused by Teredo [1] and 6to4 relays [2]. ▶ Even in situations where relays work, Teredo / 6to4 add noticeable latency [3, 4]. ▶ Tiese transition mechanisms have declined over the years due to efgorts such as − ▶ Consequentely, failure rates over IPv6 [7] have dropped signifjcantly −

Introduction Motivation Tiese HE timer values are arbitrarily chosen. What is the right timer value? [since OS X 10.11 / iOS 9] 2015 Safari uses 25 ms + history of witnessed latencies [12]. [since v15] 2012 Firefox uses parallel TCP connections [11]. [since v12.10] 2012 Opera uses parallel TCP connections [10]. [since OS X 10.7] 2011 Safari uses history of witnessed latencies [9]. [since v11] 2011 Chrome uses 300 ms [8]. Introduction | Browser Implementations Takeway Limitations Measurement Setup Browser Implementations Research Contributions Methodology Metrics and Implementation Selection of Websites Measurement Trial Lowering HE Timer Data Analysis Trends Who connects faster? Preference Slowness 3 / 18 Fragmentation of HE is visible in browser implementations today − Firefox [network.http.fast-fallback-to-IPv4=false] uses 250 ms.

Introduction Slowness 5. Lowering HE (150 ms) gives a margin benefjt of 10% and retains same preference levels. 3. HE (300 ms) makes 99% of websites prefer IPv6 more than 98% of the time. 2. 18% of websites are faster over IPv6 with 91% being at most 1 ms slower (May ’16). 1. TCP connect times to websites over IPv6 have considerably improved over time. We measure against ALEXA top 10K websites for 3 years (2013 - 2016) Introduction | Research Contributions Takeway Limitations Lowering HE Timer Preference Motivation Who connects faster? Trends Data Analysis Measurement Trial Measurement Setup Selection of Websites Metrics and Implementation Methodology Research Contributions Browser Implementations 4 / 18 4. Slower IPv6 connections are preferred in ∼ 90% of the cases.

Introduction Who connects faster? 1 Please see previous work [13] for a more detailed description of our methodology Methodology 1 Takeway Limitations Lowering HE Timer Slowness Preference Trends Motivation Data Analysis Measurement Trial Measurement Setup Selection of Websites Metrics and Implementation Methodology Research Contributions Browser Implementations 5 / 18

Introduction Who connects faster? Motivation Methodology | Metrics and Implementation Takeway Limitations Lowering HE Timer Slowness Preference 6 / 18 Trends Data Analysis Measurement Trial Measurement Setup Selection of Websites Metrics and Implementation Methodology Research Contributions Browser Implementations ▶ Uses getaddrinfo(…) to resolve service names. ▶ Uses non-blocking TCP connect(…) calls. ▶ DNS resolution time is not accounted. connection establishment times (µs) ▶ Can read multiple service names as arguments. 1) endpoint 1) service name 2) endpoint happy ▶ Can read service names list from a fjle. 2) port 3) endpoint ... n) endpoint ▶ File locking capability. happy.vaibhavbajpai.com. ▶ Sets a delay between connect(…) ; avoids SYN fmoods. ▶ Can produce both human-readable & CSV output. ▶ Cross-compiled for OpenWrt; Running on SamKnows. % happy -q 1 -m www.google.com www.facebook.com HAPPY.0;1360681039;OK;www.google.com;80;173.194.69.105;8626 HAPPY.0;1360681039;OK;www.google.com;80;2a00:1450:4008:c01::69;8884

Introduction Who connects faster? measurement targets [13]. Methodology | Selection of Websites Takeway Limitations Motivation Slowness Preference Lowering HE Timer Trends Methodology Data Analysis Research Contributions Browser Implementations Metrics and Implementation Selection of Websites Measurement Setup Measurement Trial 7 / 18 1. www.google.com 2. www.facebook.com 3. www.youtube.com 4. www.yahoo.com ▶ We use the ALEXA top 10K websites as 5. www.wikipedia.org 6. www.qq.com 7. www.blogspot.com 8. …

Introduction Trends Motivation Takeway Limitations Lowering HE Timer Slowness Preference Who connects faster? Methodology | Measurement Setup Data Analysis Research Contributions Measurement Trial Browser Implementations 8 / 18 Methodology Measurement Setup Selection of Websites Metrics and Implementation ALEXA Top 10K Websites TCP connect(...) IPv6 happy Tie happy test repeats every hour. IPv4 results DSL/Cable SamKnows Modem Tests Probe HTTPS POST Data Collector

Introduction Who connects faster? We measure from 80 dual-stacked SamKnows [14] probes. Motivation Methodology | Measurement Trial Takeway Limitations Lowering HE Timer Slowness Preference 9 / 18 Trends Data Analysis Measurement Trial Measurement Setup Selection of Websites Metrics and Implementation Methodology Research Contributions Browser Implementations NETWORK TYPE # RESIDENTIAL 55 NREN / RESEARCH 11 BUSINESS / DATACENTER 09 OPERATOR LAB 04 IXP 01 RIR # RIPE 42 ARIN 29 APNIC 07 AFRINIC 01 LACNIC 01

Introduction Motivation Data Analysis Takeway Limitations Lowering HE Timer Slowness Preference Who connects faster? Trends Data Analysis Measurement Trial Measurement Setup Selection of Websites Metrics and Implementation Methodology Research Contributions Browser Implementations 10 / 18 [2013 - 2016]

Introduction Who connects faster? Motivation Data Analysis | Trends (2013 - 2016) Takeway Limitations Lowering HE Timer Slowness Preference 11 / 18 Trends Measurement Trial Measurement Setup Selection of Websites Metrics and Implementation Methodology Research Contributions Data Analysis Browser Implementations ∆ s a ( u ) = t 4 ( u ) − t 6 ( u ) where t ( u ) is the time taken to establish TCP connection to website u . 50 0 −50 www.bing.com www.wikipedia.org −100 www.facebook.com www.youtube.com ∆sa (ms) −150 20 0 −20 www.blogspot.* www.netflix.com −40 www.yahoo.com www.google.* −60 02 05 08 11 02 05 08 11 02 05 08 11 02 05 2013 2014 2015 2016 ▶ TCP connect times to popular websites over IPv6 have considerably improved over time.

Introduction Who connects faster? Motivation ALEXA top 10K websites (as of May 2016): Data Analysis | Who connects faster? Takeway Limitations Lowering HE Timer Slowness Preference 12 / 18 Trends Measurement Trial Measurement Setup Selection of Websites Metrics and Implementation Methodology Research Contributions Data Analysis Browser Implementations 1.0 facebook 0.8 microsoft [05/2016] ▶ 18% are faster over IPv6. wikipedia 0.6 CDF youtube linkedin 0.4 google ▶ 91% of the rest are at most 1 ms slower. netflix 0.2 ALEXA (10K) ▶ 3% are at least 10 ms slower. yahoo 0.0 −1.0 −0.5 0.0 0.5 1.0 ▶ 1% are at least 100 ms slower. ∆sa (ms) ∆ s a ( u ) = t 4 ( u ) − t 6 ( u )

Introduction Preference Motivation IPv6 98% of time. 2% chance for IPv4. timer value > 300 ms Data Analysis | Preference Takeway Limitations Lowering HE Timer Slowness 13 / 18 Who connects faster? Browser Implementations Data Analysis Measurement Trial Measurement Setup Selection of Websites Metrics and Implementation Methodology Research Contributions Trends 1.0 ['13 - '16] IPv6 (189K) 0.8 300 ms ▶ Only ∼ 1% of samples above HE IPv4 (189K) CDF 0.6 0.4 0.2 0.0 10 -1 10 0 10 1 10 2 10 3 10 4 TCP Connect Times (ms) Preference (300 ms) 1.0 ['13 - '16] ▶ A 300 ms HE timer value leaves 0.8 CCDF 0.6 PROBES (80) 0.4 0.2 ALEXA (10K) ▶ 99% of top 10K ALEXA prefer 0.0 96% 97% 98% 99% 100%