INFERRING PERSISTENT INTERDOMAIN CONGESTION Amogh Dhamdhere with David Clark, Alex Gamero-Garrido, Matthew Luckie, Ricky Mok, Gautam Akiwate, Kabir Gogia, Vaibhav Bajpai, Alex Snoeren, k Claffy w w w . cai da. or
Problem: High Volume Content Strains Internet Technology and Economics 2 w w w . cai da. or
Manifestation: Interdomain Congestion Access Transit Content ISP A E B C D Content F G 3 w w w . cai da. or
Manifestation: Interdomain Congestion Access Transit Content ISP A E B Interconnection disputes resulted in congestion C D Content F G 3 w w w . cai da. or
Manifestation: Interdomain Congestion Access Transit Content ISP A E B Interconnection disputes resulted in congestion C No data was available for D third-party to study these links Content F G 3 w w w . cai da. or
Consequences of the Problem 4 w w w . cai da. or
Consequences of the Problem • Congestion on transit links affects parties other than those involved in the dispute 4 w w w . cai da. or
Consequences of the Problem • Congestion on transit links affects parties other than those involved in the dispute • Limited data available to regulators and researchers to increase transparency and empirical grounding of debate 4 w w w . cai da. or
Consequences of the Problem • Congestion on transit links affects parties other than those involved in the dispute • Limited data available to regulators and researchers to increase transparency and empirical grounding of debate • Our goal: third-party inference of congestion at interdomain interconnections 4 w w w . cai da. or
Consequences of the Problem • Congestion on transit links affects parties other than those involved in the dispute • Limited data available to regulators and researchers to increase transparency and empirical grounding of debate • Our goal: third-party inference of congestion at interdomain interconnections • Scientific approach to achieving this goal involves challenges in network inference, system development and data mining 4 w w w . cai da. or
Our Contributions 5 w w w . cai da. or
Our Contributions 1. Methodology : Operationalized a lightweight method for third-party inference of interdomain congestion, conducted thorough validation 5 w w w . cai da. or
Our Contributions 1. Methodology : Operationalized a lightweight method for third-party inference of interdomain congestion, conducted thorough validation 2. System : Built data collection and analysis platform to support the entire scientific workflow, and enable others to access and further study the data (ongoing) 5 w w w . cai da. or
Our Contributions 1. Methodology : Operationalized a lightweight method for third-party inference of interdomain congestion, conducted thorough validation 2. System : Built data collection and analysis platform to support the entire scientific workflow, and enable others to access and further study the data (ongoing) 3. Observations : Studied 8 large U.S. broadband providers from March 2016 to Dec 2017 (data collection ongoing) 5 w w w . cai da. or
Method: Time Series Latency Probes (But first, an observation) Peak-hour congestion fills up router buffers, resulting in elevated latency across an interdomain link 6 w w w . cai da. or
Method: Time Series Latency Probes (But first, an observation) Peak-hour congestion fills up router buffers, resulting in elevated latency across an interdomain link How do we measure latency across an interdomain link? 6 w w w . cai da. or
Time Series Latency Probes (TSLP) ISP A ISP B near VP far dst #A #B Border vantage point destination routers 7 w w w . cai da. or
Time Series Latency Probes (TSLP) ISP A ISP B near VP far dst #A #B Border vantage point destination routers TTL: n RTT #A 7 w w w . cai da. or
Time Series Latency Probes (TSLP) ISP A ISP B near VP far dst #A #B Border vantage point destination routers TTL: n RTT #A TTL: n+1 RTT #B 7 w w w . cai da. or
Time Series Latency Probes (TSLP) ISP A ISP B near VP far dst #A #B Border vantage point destination routers TTL: n RTT #A TTL: n+1 RTT #B (repeat to obtain a time series) 7 w w w . cai da. or
An Experiment with TSLP Comcast Cogent near VP far dst #A #B Border vantage point destination routers Measured interdomain link from Comcast to Cogent using VP in Comcast 8 w w w . cai da. or
An Experiment with TSLP RTT measurements of border routers 120 Cogent (far) Comcast (near) 100 RTT (ms) 80 60 40 20 0 Thu Fri Sat Sun Mon Tue Wed Thu 7th 8th 9th 10th 11th 12th 13th 14th Day of week (local time in New York) *Luckie, Dhamdhere, Clark, Huffaker, Claffy, “Challenges 9 in Inferring Interdomain Congestion”, IMC 2014 w w w . cai da. or
An Experiment with TSLP RTT measurements of border routers 120 Cogent (far) Comcast (near) 100 RTT (ms) 80 60 40 20 0 Thu Fri Sat Sun Mon Tue Wed Thu 7th 8th 9th 10th 11th 12th 13th 14th Day of week (local time in New York) Diurnal elevation to far-side *Luckie, Dhamdhere, Clark, Huffaker, Claffy, “Challenges 9 in Inferring Interdomain Congestion”, IMC 2014 w w w . cai da. or
An Experiment with TSLP RTT measurements of border routers 120 Cogent (far) Comcast (near) 100 RTT (ms) 80 60 40 20 0 Thu Fri Sat Sun Mon Tue Wed Thu 7th 8th 9th 10th 11th 12th 13th 14th Day of week (local time in New York) Diurnal elevation to far-side No elevation to near-side *Luckie, Dhamdhere, Clark, Huffaker, Claffy, “Challenges 10 in Inferring Interdomain Congestion”, IMC 2014 w w w . cai da. or
Pieces of the Puzzle 11 w w w . cai da. or
Pieces of the Puzzle 12 w w w . cai da. or
Pieces of the Puzzle • Interdomain link Identification: Need to identify interdomain links to be able to probe them 12 w w w . cai da. or
Pieces of the Puzzle • Interdomain link Identification: Need to identify interdomain links to be able to probe them • Adaptive Probing: Need to be adaptive to changes in the underlying topology and routing 12 w w w . cai da. or
Pieces of the Puzzle • Interdomain link Identification: Need to identify interdomain links to be able to probe them • Adaptive Probing: Need to be adaptive to changes in the underlying topology and routing • Identifying Congested Links: Need time-series analysis techniques to find patterns in (noisy) data that indicate congestion 12 w w w . cai da. or
Pieces of the Puzzle • Interdomain link Identification: Need to identify interdomain links to be able to probe them • Adaptive Probing: Need to be adaptive to changes in the underlying topology and routing • Identifying Congested Links: Need time-series analysis techniques to find patterns in (noisy) data that indicate congestion • Validation: Need to validate inferences. Most peering agreements are covered by NDAs 12 w w w . cai da. or
System bdrmap External bdrmap analysis Inputs probing Interactive Data Exploration Links TSLP DB Probing Real-time history Dashboards TSLP target selection Longitudinal Youtube NDT Views Loss rate Frontend TSLP VP Time series data analysis Backend M easurement and AN alysis of I nternet 13 C ongestion w w w . cai da. or
System Interdomain link identification bdrmap External bdrmap analysis Inputs probing Interactive Data Exploration Links TSLP DB Probing Real-time history Dashboards TSLP target selection Longitudinal Youtube NDT Views Loss rate Frontend TSLP VP Time series data analysis Backend M easurement and AN alysis of I nternet 13 C ongestion w w w . cai da. or
System bdrmap External bdrmap analysis Inputs probing Interactive Data Exploration Links TSLP DB Probing Real-time history Dashboards TSLP target Adaptive selection Longitudinal Probing Youtube NDT Views Loss rate Frontend TSLP VP Time series data analysis Backend M easurement and AN alysis of I nternet 13 C ongestion w w w . cai da. or
System bdrmap External bdrmap analysis Inputs probing Interactive Data Exploration Links TSLP DB Probing Real-time history Dashboards TSLP target selection Longitudinal Youtube NDT Views Loss Identifying Congested Links rate Frontend TSLP VP Time series data analysis Backend M easurement and AN alysis of I nternet 13 C ongestion w w w . cai da. or
System bdrmap External bdrmap analysis Inputs probing Interactive Data Exploration Links TSLP DB Probing Real-time history Dashboards Validation TSLP target selection Longitudinal Youtube NDT Views Loss rate Frontend TSLP VP Time series data analysis Backend M easurement and AN alysis of I nternet 13 C ongestion w w w . cai da. or
System bdrmap External bdrmap analysis Inputs probing Interactive Data Exploration Links TSLP DB Probing Real-time history Dashboards TSLP target selection Longitudinal Youtube NDT Views Loss rate Frontend TSLP VP Time series data analysis Visualization Backend M easurement and AN alysis of I nternet 13 C ongestion w w w . cai da. or
Identifying Interdomain Links ISP A E B VP C D F G 14 w w w . cai da. or
Recommend
More recommend
