Overview Measuring IPv6 Performance Nov 25, 2016 TU Munich, Germany Prof. Dr. Jörg Ott Aalto University, Finland Saba Ahsan SamKnows Limited, London, UK Steffje Jacob Eravuchira | Sam Crawford Jacobs University Bremen, Germany Prof. Dr. Jürgen Schönwälder Joint Work with Raitenhaslach, Germany Munich Internet Research Retreat, Jacobs University, Bremen Vaibhav Bajpai Q/A TCP connect times Takeway Causality Analysis Success Rate Web Similarity Lowering HE Timer Slowness Preference Happy Eyeballs Preference Latency YouTube Preference Who connects faster? Trends 1 / 27
Overview Slowness Tiis research would not have been possible without these amazing people! Q/A Takeway Causality Analysis TCP connect times Web Similarity Lowering HE Timer Success Rate Preference Who connects faster? Happy Eyeballs Trends 2 / 27 Preference YouTube Latency Preference • • – – – – – • What’s ¡missing: ¡Many ¡things, ¡but ¡in
Overview Takeway Bajpai et al. [4] Eravuchira et al. [3] Bajpai et al. [2] Ahsan et al. [1] Measuring IPv6 Performance Web Similarity YouTube Happy Eyeballs TCP Connect Times Overview Q/A Causality Analysis TCP connect times Success Rate Web Similarity Lowering HE Timer Slowness Preference Happy Eyeballs Preference Latency YouTube Preference Who connects faster? Trends 3 / 27
Overview Success Rate ARIN LACNIC RIPE APNIC TCP connect times Overview | Motivation Q/A Takeway Causality Analysis 4 / 27 Web Similarity Lowering HE Timer Slowness Preference Happy Eyeballs Preference Latency Trends Who connects faster? Preference YouTube ▶ Literature has largely focussed on measuring IPv6 adoption [5, 6, 7] ( ′ 10 − ′ 14). ▶ Addressing ▶ Naming ▶ Routing ▶ Reachability ▶ Very little work [8] on measuring performance of service delivery over IPv6. ▶ Largely due to lack of available content over IPv6. ▶ A number of signifjcant events occured during the span of this dissertation. Apr ′ 11 ▶ IANA IPv4 Address Exhaustion [9] ▶ World IPv6 Day ′ 11 [10] Sep ′ 12 ▶ World IPv6 Launch Day ′ 12 [11] Jun ′ 14 ▶ RIR IPv4 Address Exhaustion [9] Sep ′ 15
Overview Q/A 1 Comcast, Deutsche Telekom AG, AT&T, Verizon Wireless, T-Mobile USA 26.61% Germany 26.95% Switzerland 30.12% United States 47.38% Belgium 11/2016 12.46% 09/2012 0.85% TCP connect times Overview | Motivation Takeway Causality Analysis Trends Who connects faster? Preference YouTube Latency Preference Happy Eyeballs Preference Slowness Lowering HE Timer Web Similarity Success Rate 5 / 27 ▶ Large IPv6 broadband rollouts 1 [4]. ▶ Global IPv6 adoption [12]. ▶ Tiis study closes the gap. ▶ It measures IPv6 performance of operational dual-stacked content delivery services.
Overview Lowering HE Timer We measure from 80 dual-stacked SamKnows [13] probes. TCP connect times Overview | Measurement Trial Q/A Takeway Causality Analysis Success Rate Web Similarity 6 / 27 Slowness Preference Happy Eyeballs Preference Latency YouTube Preference Who connects faster? Trends 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
Overview Takeway * entries are papers currently under review. Bajpai et al. [4] Eravuchira et al. [3] Bajpai et al. [2] Ahsan et al. [1] Measuring IPv6 Performance Web Similarity YouTube Happy Eyeballs TCP Connect Times Overview | TCP Connect Times Q/A Causality Analysis TCP connect times Success Rate Web Similarity Lowering HE Timer Slowness Preference Happy Eyeballs Preference Latency YouTube Preference Who connects faster? Trends 7 / 27
Overview Lowering HE Timer TCP connect times TCP Connect Times | Trends (2013 - 2016) Q/A Takeway Causality Analysis Success Rate Web Similarity 8 / 27 Slowness Happy Eyeballs Preference Latency YouTube Preference Who connects faster? Preference Trends ∆ 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.
Overview Lowering HE Timer TCP connect times ALEXA top 10K websites (as of May 2016): TCP Connect Times | Who connects faster? Q/A Takeway Causality Analysis Success Rate Web Similarity 9 / 27 Slowness Happy Eyeballs Preference Latency YouTube Preference Who connects faster? Preference Trends 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 )
Overview Lowering HE Timer TCP connect times TCP Connect Times | IPv6 Preference Q/A Takeway Causality Analysis Success Rate Web Similarity 10 / 27 Slowness Preference Happy Eyeballs Preference Latency YouTube Preference Who connects faster? Trends Preference (300 ms) [2013 - 2016] 1.0 0.8 CCDF 0.6 PROBES (80) 0.4 ALEXA (10K) 0.2 0.0 96% 97% 98% 99% 100% ▶ A 300 ms HE timer value leaves 2% chance for IPv4. ▶ 99% of top 10K ALEXA prefer IPv6 98% of time.
Overview Takeway * entries are papers currently under review. Bajpai et al. [4] Eravuchira et al. [3] Bajpai et al. [2] Ahsan et al. [1] Measuring IPv6 Performance Web Similarity YouTube Happy Eyeballs TCP Connect Times Overview | Measuring YouTube Q/A Causality Analysis TCP connect times Success Rate Web Similarity Lowering HE Timer Slowness Preference Happy Eyeballs Preference Latency YouTube Preference Who connects faster? Trends 11 / 27
Overview Lowering HE Timer TCP connect times YouTube | Latency Q/A Takeway Causality Analysis Success Rate Web Similarity 12 / 27 Slowness Preference Trends Who connects faster? Preference YouTube Latency Preference Happy Eyeballs TCP Connect Times 0 ∆t (ms) −1 −2 −3 Web −4 ▶ TCP connect times −5 TCP Connect Times 0.0 ▶ < 1 ms slower over IPv6 ∆t (ms) −0.1 Audio −0.2 Video ▶ Higher towards webpages −0.3 −0.4 Prebuffering Duration ▶ Prebufgering durations ∆p (ms) 0 −40 −80 ▶ > 25 ms slower over IPv6 −120 Startup Delay 0 ∆s (ms) −100 ▶ Startup delay −200 −300 −400 ▶ > 100 ms slower over IPv6 Oct Jan Apr Jul Oct Jan Apr 2015 2016 ∆ t ( y ) = tc 4 ( y ) − tc 6 ( y ) ∆ p ( y ) = pd 4 ( y ) − pd 6 ( y ) ∆ s ( y ) = sd 4 ( y ) − sd 6 ( y ) Latency is consistently higher over IPv6.
Overview Lowering HE Timer TCP connect times YouTube | IPv6 Preference Q/A Takeway Causality Analysis Success Rate Web Similarity 13 / 27 Slowness Preference Happy Eyeballs Preference Latency YouTube Preference Who connects faster? Trends IPv6 Preference 1.0 ['14 - '16] 0.8 Web (458) 0.6 CCDF Audio (458) 0.4 0.2 Video (458) 0.0 95% 96% 97% 98% 99% 100% ▶ Media streams are preferred over IPv6 more than 97% of the time.
Overview Takeway * entries are papers currently under review. Bajpai et al. [4] Eravuchira et al. [3] Bajpai et al. [2] Ahsan et al. [1] Measuring IPv6 Performance Web Similarity YouTube Happy Eyeballs TCP Connect Times Overview | Measuring Efgects of Happy Eyeballs Q/A Causality Analysis TCP connect times Success Rate Web Similarity Lowering HE Timer Slowness Preference Happy Eyeballs Preference Latency YouTube Preference Who connects faster? Trends 14 / 27
Overview Lowering HE Timer TCP connect times Happy Eyeballs | Preference Q/A Takeway Causality Analysis Success Rate Web Similarity 15 / 27 Slowness Trends Happy Eyeballs Preference Latency YouTube Preference Who connects faster? Preference 1.0 [2013 - 2016] IPv6 (89K) 0.8 300 ms IPv4 (89K) 0.6 CDF 0.4 0.2 0.0 10 -1 10 0 10 1 10 2 10 3 10 4 TCP Connect Times (ms) ▶ Only ∼ 1% of samples above HE timer value > 300 ms
Overview Web Similarity Can a lower HE timer provide same preference over IPv6 but not penalise IPv4 when it’s faster? TCP connect times apart from IPv4 connections 90% of the time. Happy Eyeballs | Slowness Q/A Takeway Causality Analysis Success Rate 16 / 27 Lowering HE Timer Preference Preference Happy Eyeballs Preference Latency Slowness Who connects faster? YouTube Trends 1.0 [2013 - 2016] 93% 89K Samples where HE prefers IPv6 − 0.8 99% 0.6 CDF 30% 0.4 ▶ HE prefers slower IPv6 1% 7% 2% 0.2 0.0 ▶ Absolute difgerence is not that far −40 −30 −20 −10 0 10 ∆sa (ms) ▶ 30% − at least 1 ms slower. ▶ 7% − at least 10 ms slower. ∆ s a ( u ) = t 4 ( u ) − t 6 ( u ) ∆ s r ( u ) = t 4( u ) − t 6( u ) t 4( u )
Recommend
More recommend
