INFERRING PERSISTENT INTERDOMAIN CONGESTION Amogh Dhamdhere with - - PowerPoint PPT Presentation

▶
inferring persistent interdomain congestion
SMART_READER_LITE
LIVE PREVIEW

INFERRING PERSISTENT INTERDOMAIN CONGESTION Amogh Dhamdhere with - - PowerPoint PPT Presentation

Jan 07, 2024 185 likes •1.04k views

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

slide-1
SLIDE 1

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.

  • r
slide-2
SLIDE 2

w w w . cai da.

  • r

2

Problem: High Volume Content Strains Internet Technology and Economics

slide-3
SLIDE 3

w w w . cai da.

  • r

Manifestation: Interdomain Congestion

3

C B ISP A D E F G Access Content Content Transit

slide-4
SLIDE 4

w w w . cai da.

  • r

Manifestation: Interdomain Congestion

3

C B ISP A Interconnection disputes resulted in congestion D E F G Access Content Content Transit

slide-5
SLIDE 5

w w w . cai da.

  • r

Manifestation: Interdomain Congestion

3

C B ISP A Interconnection disputes resulted in congestion D E F G Access Content Content Transit No data was available for third-party to study these links

slide-6
SLIDE 6

w w w . cai da.

  • r

Consequences of the Problem

4

slide-7
SLIDE 7

w w w . cai da.

  • r

Consequences of the Problem

  • Congestion on transit links affects parties other than those

involved in the dispute

4

slide-8
SLIDE 8

w w w . cai da.

  • r

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

slide-9
SLIDE 9

w w w . cai da.

  • r

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

slide-10
SLIDE 10

w w w . cai da.

  • r

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

slide-11
SLIDE 11

w w w . cai da.

  • r

Our Contributions

5

slide-12
SLIDE 12

w w w . cai da.

  • r

Our Contributions

  • 1. Methodology: Operationalized a lightweight method for

third-party inference of interdomain congestion, conducted thorough validation

5

slide-13
SLIDE 13

w w w . cai da.

  • r

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

slide-14
SLIDE 14

w w w . cai da.

  • r

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

slide-15
SLIDE 15

w w w . cai da.

  • r

Method: Time Series Latency Probes

6

Peak-hour congestion fills up router buffers, resulting in elevated latency across an interdomain link

(But first, an observation)

slide-16
SLIDE 16

w w w . cai da.

  • r

Method: Time Series Latency Probes

6

Peak-hour congestion fills up router buffers, resulting in elevated latency across an interdomain link How do we measure latency across an interdomain link?

(But first, an observation)

slide-17
SLIDE 17

w w w . cai da.

  • r

Time Series Latency Probes (TSLP)

7

ISP B ISP A near far dst VP Border routers #A #B vantage point destination

slide-18
SLIDE 18

w w w . cai da.

  • r

Time Series Latency Probes (TSLP)

7

ISP B ISP A near far dst VP Border routers #A #B vantage point destination RTT #A

TTL: n

slide-19
SLIDE 19

w w w . cai da.

  • r

Time Series Latency Probes (TSLP)

7

ISP B ISP A near far dst VP Border routers #A #B vantage point destination RTT #A

TTL: n

RTT #B

TTL: n+1

slide-20
SLIDE 20

w w w . cai da.

  • r

Time Series Latency Probes (TSLP)

7

ISP B ISP A near far dst VP Border routers #A #B vantage point destination (repeat to obtain a time series) RTT #A

TTL: n

RTT #B

TTL: n+1

slide-21
SLIDE 21

w w w . cai da.

  • r

8

Measured interdomain link from Comcast to Cogent using VP in Comcast

Comcast near far dst VP Border routers #A #B vantage point destination

An Experiment with TSLP

Cogent

slide-22
SLIDE 22

w w w . cai da.

  • r

9 *Luckie, Dhamdhere, Clark, Huffaker, Claffy, “Challenges in Inferring Interdomain Congestion”, IMC 2014

An Experiment with TSLP

RTT measurements of border routers Day of week (local time in New York) 40 60 80 100 120 RTT (ms) Comcast (near) Cogent (far) Thu 7th Fri 8th Sat 9th Sun 10th Mon 11th Tue 12th Wed 13th Thu 14th 20

slide-23
SLIDE 23

w w w . cai da.

  • r

9 *Luckie, Dhamdhere, Clark, Huffaker, Claffy, “Challenges in Inferring Interdomain Congestion”, IMC 2014

An Experiment with TSLP

RTT measurements of border routers Day of week (local time in New York) 40 60 80 100 120 RTT (ms) Comcast (near) Cogent (far) Thu 7th Fri 8th Sat 9th Sun 10th Mon 11th Tue 12th Wed 13th Thu 14th 20

Diurnal elevation to far-side

slide-24
SLIDE 24

w w w . cai da.

  • r

10

RTT measurements of border routers 100 Thu 7th Fri 8th Sat 9th Sun 10th Mon 11th Tue 12th Wed 13th Thu 14th Day of week (local time in New York) 120 RTT (ms) Comcast (near) 20 40 60 80 Cogent (far)

An Experiment with TSLP

*Luckie, Dhamdhere, Clark, Huffaker, Claffy, “Challenges in Inferring Interdomain Congestion”, IMC 2014

Diurnal elevation to far-side No elevation to near-side

slide-25
SLIDE 25

w w w . cai da.

  • r

Pieces of the Puzzle

11

slide-26
SLIDE 26

w w w . cai da.

  • r

Pieces of the Puzzle

12

slide-27
SLIDE 27

w w w . cai da.

  • r

Pieces of the Puzzle

  • Interdomain link Identification: Need to identify

interdomain links to be able to probe them

12

slide-28
SLIDE 28

w w w . cai da.

  • r

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

slide-29
SLIDE 29

w w w . cai da.

  • r

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

slide-30
SLIDE 30

w w w . cai da.

  • r

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

slide-31
SLIDE 31

w w w . cai da.

  • r

13

VP

bdrmap analysis

Backend

Links DB Probing history TSLP data

Frontend

Real-time Dashboards Longitudinal Views Time series analysis bdrmap probing TSLP NDT Loss rate Youtube Interactive Data Exploration TSLP target selection External Inputs

System

Measurement and ANalysis of Internet Congestion

slide-32
SLIDE 32

w w w . cai da.

  • r

13

VP

bdrmap analysis

Backend

Links DB Probing history TSLP data

Frontend

Real-time Dashboards Longitudinal Views Time series analysis bdrmap probing TSLP NDT Loss rate Youtube Interactive Data Exploration TSLP target selection External Inputs

System

Interdomain link identification

Measurement and ANalysis of Internet Congestion

slide-33
SLIDE 33

w w w . cai da.

  • r

13

VP

bdrmap analysis

Backend

Links DB Probing history TSLP data

Frontend

Real-time Dashboards Longitudinal Views Time series analysis bdrmap probing TSLP NDT Loss rate Youtube Interactive Data Exploration TSLP target selection External Inputs

System

Measurement and ANalysis of Internet Congestion

Adaptive Probing

slide-34
SLIDE 34

w w w . cai da.

  • r

13

VP

bdrmap analysis

Backend

Links DB Probing history TSLP data

Frontend

Real-time Dashboards Longitudinal Views Time series analysis bdrmap probing TSLP NDT Loss rate Youtube Interactive Data Exploration TSLP target selection External Inputs

System

Measurement and ANalysis of Internet Congestion

Identifying Congested Links

slide-35
SLIDE 35

w w w . cai da.

  • r

13

VP

bdrmap analysis

Backend

Links DB Probing history TSLP data

Frontend

Real-time Dashboards Longitudinal Views Time series analysis bdrmap probing TSLP NDT Loss rate Youtube Interactive Data Exploration TSLP target selection External Inputs

System

Measurement and ANalysis of Internet Congestion

Validation

slide-36
SLIDE 36

w w w . cai da.

  • r

13

VP

bdrmap analysis

Backend

Links DB Probing history TSLP data

Frontend

Real-time Dashboards Longitudinal Views Time series analysis bdrmap probing TSLP NDT Loss rate Youtube Interactive Data Exploration TSLP target selection External Inputs

System

Measurement and ANalysis of Internet Congestion

Visualization

slide-37
SLIDE 37

w w w . cai da.

  • r

Identifying Interdomain Links

14

B ISP A D E C F G VP

slide-38
SLIDE 38

w w w . cai da.

  • r

Identifying Interdomain Links

15

E We focus on interdomain links of network hosting a measurement VP C B ISP A D F G VP

slide-39
SLIDE 39

w w w . cai da.

  • r

Identifying Interdomain Links

16

C B ISP A D E F G

*Luckie, Dhamdhere, Huffaker, Clark, Claffy, “bdrmap: Inference of Borders Between IP Networks”, IMC 2016

slide-40
SLIDE 40

w w w . cai da.

  • r

Identifying Interdomain Links

16

bdrmap: traceroute to all routed IPv4 prefixes C B ISP A D E F G

*Luckie, Dhamdhere, Huffaker, Clark, Claffy, “bdrmap: Inference of Borders Between IP Networks”, IMC 2016

slide-41
SLIDE 41

w w w . cai da.

  • r

Identifying Interdomain Links

17

C B ISP A D E F G

*Luckie, Dhamdhere, Huffaker, Clark, Claffy, “bdrmap: Inference of Borders Between IP Networks”, IMC 2016

slide-42
SLIDE 42

w w w . cai da.

  • r

Identifying Interdomain Links

17

bdrmap: Infer router ownership to identify interdomain links at the router-level C B ISP A D E F G

*Luckie, Dhamdhere, Huffaker, Clark, Claffy, “bdrmap: Inference of Borders Between IP Networks”, IMC 2016

slide-43
SLIDE 43

w w w . cai da.

  • r

Identifying Congested Links

  • Focus on persistently congested links
  • Look for periods of elevated latency that correlate across

days (autocorrelation method)

18

slide-44
SLIDE 44

w w w . cai da.

  • r

Autocorrelation method

19

Day 1 Day 2 Day 3 Day 50 Far-side latency

slide-45
SLIDE 45

w w w . cai da.

  • r

Autocorrelation method

19

Day 1 Day 2 Day 3 Day 50 Far-side latency

slide-46
SLIDE 46

w w w . cai da.

  • r

Autocorrelation method

19

Day 1 Day 2 Day 3 Day 50 Sum number of days showing elevation in each 15-min interval Far-side latency # Days

slide-47
SLIDE 47

w w w . cai da.

  • r

Autocorrelation method

20

Infer an interval of recurring congestion Day 1 Day 2 Day 3 Day 50 # Days Far-side latency

slide-48
SLIDE 48

w w w . cai da.

  • r

Autocorrelation method

21

Map inferred interval

  • nto each day

# Days Day 1 Day 2 Day 3 Day 50 Far-side latency

slide-49
SLIDE 49

w w w . cai da.

  • r

Autocorrelation method

21

Map inferred interval

  • nto each day

Reject elevated intervals that do not overlap with recurring interval # Days Day 1 Day 2 Day 3 Day 50 Far-side latency

slide-50
SLIDE 50

w w w . cai da.

  • r

Heavy Emphasis on Validation

22

(Critical due to political nature of inferences)

slide-51
SLIDE 51

w w w . cai da.

  • r

Heavy Emphasis on Validation

22

(Critical due to political nature of inferences)

  • Validation of inference method
  • Correlation with loss
  • Correlation with throughput
  • Correlation with

YouTube streaming

slide-52
SLIDE 52

w w w . cai da.

  • r

Heavy Emphasis on Validation

22

(Critical due to political nature of inferences)

  • Validation of inference method
  • Correlation with loss
  • Correlation with throughput
  • Correlation with

YouTube streaming

  • Validation of specific inferences
  • Ground truth from operators
slide-53
SLIDE 53

w w w . cai da.

  • r

Heavy Emphasis on Validation

  • Validation of inference method
  • Correlation with loss
  • Correlation with throughput
  • Correlation with

YouTube streaming

  • Validation of specific inferences
  • Ground truth from operators

23

(Critical due to political nature of inferences)

slide-54
SLIDE 54

w w w . cai da.

  • r

Heavy Emphasis on Validation

  • Validation of inference method
  • Correlation with loss
  • Correlation with throughput
  • Correlation with

YouTube streaming

  • Validation of specific inferences
  • Ground truth from operators

23

(Critical due to political nature of inferences) Brief overview in this talk See paper for full details

slide-55
SLIDE 55

w w w . cai da.

  • r

24

TSLP latency

10 20 30 40 50 60 70 80 90 RTT 15-min minimum (ms)

Far RTT Near RTT

6 : 1 8 : 6 : 1 8 : 6 : 1 8 : 6 : UTC Time 0.0

Does TSLP inference correlate with loss?

slide-56
SLIDE 56

w w w . cai da.

  • r

24

TSLP latency

Diurnal latency elevation to far side indicates congestion

10 20 30 40 50 60 70 80 90 RTT 15-min minimum (ms)

Far RTT Near RTT

6 : 1 8 : 6 : 1 8 : 6 : 1 8 : 6 : UTC Time 0.0

Does TSLP inference correlate with loss?

slide-57
SLIDE 57

w w w . cai da.

  • r

Does TSLP inference correlate with loss?

25

10 20 30 40 50 60 70 80 90 RTT 15-min minimum (ms)

Far RTT Near RTT

6 : 1 8 : 6 : 1 8 : 6 : 1 8 : 6 : UTC Time 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 Loss rate (5-min average)

Far Loss Near Loss

TSLP latency Loss Rate

slide-58
SLIDE 58

w w w . cai da.

  • r

Does TSLP inference correlate with loss?

25

Far loss rate exceeds near loss rate during periods of inferred congestion

10 20 30 40 50 60 70 80 90 RTT 15-min minimum (ms)

Far RTT Near RTT

6 : 1 8 : 6 : 1 8 : 6 : 1 8 : 6 : UTC Time 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 Loss rate (5-min average)

Far Loss Near Loss

TSLP latency Loss Rate

slide-59
SLIDE 59

w w w . cai da.

  • r

Does TSLP inference correlate with loss?

26

Far loss rate during inferred congestion exceeds far loss rate during non-congested period

10 20 30 40 50 60 70 80 90 RTT 15-min minimum (ms)

Far RTT Near RTT

6 : 1 8 : 6 : 1 8 : 6 : 1 8 : 6 : UTC Time 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 Loss rate (5-min average)

Far Loss Near Loss

TSLP latency Loss Rate

slide-60
SLIDE 60

w w w . cai da.

  • r

Does TSLP inference correlate with loss?

27

In 81% of cases, far loss rate was higher during congested periods and higher than near loss rate

10 20 30 40 50 60 70 80 90 RTT 15-min minimum (ms)

Far RTT Near RTT

6 : 1 8 : 6 : 1 8 : 6 : 1 8 : 6 : UTC Time 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 Loss rate (5-min average)

Far Loss Near Loss

TSLP latency Loss Rate

slide-61
SLIDE 61

w w w . cai da.

  • r

Does TSLP inference correlate with throughput?

28

M-lab NDT server

congested link

Approach: throughput measurements from Ark VP to M-lab NDT server traversing congested interdomain link ISP A ISP B

Ark VP

“near” side “far” side

slide-62
SLIDE 62

w w w . cai da.

  • r

Does TSLP inference correlate with throughput?

28

M-lab NDT server

congested link

Approach: throughput measurements from Ark VP to M-lab NDT server traversing congested interdomain link ISP A ISP B

Ark VP

“near” side “far” side

Challenge: difficult to find NDT servers that cover specifically observed interconnections

slide-63
SLIDE 63

w w w . cai da.

  • r

Does TSLP inference correlate with throughput?

29

15 20 25 30 35 40 45 RTT 15-min minimums (ms)

Far RTT Near RTT

06:00 18:00 06:00 18:00 06:00 18:00 06:00 18:00 06:00 UTC Time 5 10 15 20 25 30 35 40 Throughput (Mbps)

Download Throughput

slide-64
SLIDE 64

w w w . cai da.

  • r

Does TSLP inference correlate with throughput?

Lower throughput during periods inferred congested

29

15 20 25 30 35 40 45 RTT 15-min minimums (ms)

Far RTT Near RTT

06:00 18:00 06:00 18:00 06:00 18:00 06:00 18:00 06:00 UTC Time 5 10 15 20 25 30 35 40 Throughput (Mbps)

Download Throughput

slide-65
SLIDE 65

w w w . cai da.

  • r

Does TSLP inference correlate with throughput?

  • Avg. throughput 27Mbps

during uncongested periods

30

  • Avg. throughput 8Mbps

during congested periods

15 20 25 30 35 40 45 RTT 15-min minimums (ms)

Far RTT Near RTT

06:00 18:00 06:00 18:00 06:00 18:00 06:00 18:00 06:00 UTC Time 5 10 15 20 25 30 35 40 Throughput (Mbps)

Download Throughput

slide-66
SLIDE 66

w w w . cai da.

  • r

Validation: Operator Feedback

31

slide-67
SLIDE 67

w w w . cai da.

  • r

Validation: Operator Feedback

  • Validated our inferences with operators from two large

U.S. access ISPs

31

slide-68
SLIDE 68

w w w . cai da.

  • r

Validation: Operator Feedback

  • Validated our inferences with operators from two large

U.S. access ISPs

  • ISP A: 7 links (all inferred congested)
  • ISP B: 20 links (10 inferred congested, 10 uncongested)

31

slide-69
SLIDE 69

w w w . cai da.

  • r

Validation: Operator Feedback

  • Validated our inferences with operators from two large

U.S. access ISPs

  • ISP A: 7 links (all inferred congested)
  • ISP B: 20 links (10 inferred congested, 10 uncongested)
  • Our inferences were correct in each case: no false

positives or false negatives

31

slide-70
SLIDE 70

w w w . cai da.

  • r

Longitudinal Study

  • Collecting data since March 2016
  • Focused on interdomain links of 8 large access ISPs in the

U.S. to their transit providers and peers from Mar 2016 to Dec 2017

  • Driving questions:
  • How prevalent is interdomain congestion?
  • Which transit/content providers are most often congested to access

providers?

  • Can we characterize trends over time?

32

slide-71
SLIDE 71

w w w . cai da.

  • r

What Did We Find (so far)?

  • No evidence of widespread (pervasive) congestion

between Mar 2016 and Dec 2017

  • Small fraction of peers of the 8 studied access providers

showed evidence of congestion

  • Certain transit providers (e.g., TATA) and content providers

(e.g. Google) most often showed evidence of congestion

  • Interesting dynamics of interdomain congestion

See paper for details

33

slide-72
SLIDE 72

w w w . cai da.

  • r

34

Content Providers: Content Content Transit Transit Transit Zayo NTT Tata Transit Google Netflix Providers: Content Telia XO Vodafone Level3 Transit

Cox Verizon CenturyLink AT&T Comcast TWC

Content Transit Transit Content Content 80 100 80 60 40 60 40 100 80 60 40 20 ’17 Dec ’17 Sep ’17 Jun 20 100 80 60 40 20 20 100 80 60 40 20 100 80 60 40 20 100 80 60 40 Mar ’16 Jun ’16 Sep ’16 Dec ’16 Mar ’17 Jun ’17 Sep ’17 Dec ’17 20 20 40 60 80 100 ’17 Mar ’16 Dec ’16 Sep ’16 Jun ’16 Mar 100 80 60 40 20 100 80 60 40 20 100 80 40 60 80 100 60 40 20 100 20

Percent of congested day-links over time

slide-73
SLIDE 73

w w w . cai da.

  • r

35

Content Providers: Content Content Transit Transit Transit Zayo NTT Tata Transit Google Netflix Providers: Content Telia XO Vodafone Level3 Transit

Cox Verizon CenturyLink AT&T Comcast TWC

Content Transit Transit Content Content 80 100 80 60 40 60 40 100 80 60 40 20 ’17 Dec ’17 Sep ’17 Jun 20 100 80 60 40 20 20 100 80 60 40 20 100 80 60 40 20 100 80 60 40 Mar ’16 Jun ’16 Sep ’16 Dec ’16 Mar ’17 Jun ’17 Sep ’17 Dec ’17 20 20 40 60 80 100 ’17 Mar ’16 Dec ’16 Sep ’16 Jun ’16 Mar 100 80 60 40 20 100 80 60 40 20 100 80 40 60 80 100 60 40 20 100 20 Providers: Transit Zayo NTT Tata Google Netflix Providers: Content Telia XO Vodafone Level3 Content Transit

Comcast

100 80 60 40 20 100 80 60 40 20

Percent of congested day-links over time

’17 Dec ’17 Sep ’17 Jun ’17 Mar ’16 Dec ’16 Sep ’16 Jun ’16 Mar

slide-74
SLIDE 74

w w w . cai da.

  • r

35

Content Providers: Content Content Transit Transit Transit Zayo NTT Tata Transit Google Netflix Providers: Content Telia XO Vodafone Level3 Transit

Cox Verizon CenturyLink AT&T Comcast TWC

Content Transit Transit Content Content 80 100 80 60 40 60 40 100 80 60 40 20 ’17 Dec ’17 Sep ’17 Jun 20 100 80 60 40 20 20 100 80 60 40 20 100 80 60 40 20 100 80 60 40 Mar ’16 Jun ’16 Sep ’16 Dec ’16 Mar ’17 Jun ’17 Sep ’17 Dec ’17 20 20 40 60 80 100 ’17 Mar ’16 Dec ’16 Sep ’16 Jun ’16 Mar 100 80 60 40 20 100 80 60 40 20 100 80 40 60 80 100 60 40 20 100 20 Providers: Transit Zayo NTT Tata Google Netflix Providers: Content Telia XO Vodafone Level3

Comcast-Google congestion increased and subsided over time

Content Transit

Comcast

100 80 60 40 20 100 80 60 40 20

Percent of congested day-links over time

’17 Dec ’17 Sep ’17 Jun ’17 Mar ’16 Dec ’16 Sep ’16 Jun ’16 Mar

slide-75
SLIDE 75

w w w . cai da.

  • r

36

Content Providers: Content Content Transit Transit Transit Zayo NTT Tata Transit Google Netflix Providers: Content Telia XO Vodafone Level3 Transit

Cox Verizon CenturyLink AT&T Comcast TWC

Content Transit Transit Content Content 80 100 80 60 40 60 40 100 80 60 40 20 ’17 Dec ’17 Sep ’17 Jun 20 100 80 60 40 20 20 100 80 60 40 20 100 80 60 40 20 100 80 60 40 Mar ’16 Jun ’16 Sep ’16 Dec ’16 Mar ’17 Jun ’17 Sep ’17 Dec ’17 20 20 40 60 80 100 ’17 Mar ’16 Dec ’16 Sep ’16 Jun ’16 Mar 100 80 60 40 20 100 80 60 40 20 100 80 40 60 80 100 60 40 20 100 20 Providers: Transit Zayo NTT Tata Google Netflix Providers: Content Telia XO Vodafone Level3

Time Warner - TATA congestion subsided in December 2016

Content Transit

TWC

100 80 60 40 20 100 80 60 40 20 ’17 Dec ’17 Sep ’17 Jun ’17 Mar ’16 Dec ’16 Sep ’16 Jun ’16 Mar

slide-76
SLIDE 76

w w w . cai da.

  • r

37

Content Providers: Content Content Transit Transit Transit Zayo NTT Tata Transit Google Netflix Providers: Content Telia XO Vodafone Level3 Transit

Cox Verizon CenturyLink AT&T Comcast TWC

Content Transit Transit Content Content 80 100 80 60 40 60 40 100 80 60 40 20 ’17 Dec ’17 Sep ’17 Jun 20 100 80 60 40 20 20 100 80 60 40 20 100 80 60 40 20 100 80 60 40 Mar ’16 Jun ’16 Sep ’16 Dec ’16 Mar ’17 Jun ’17 Sep ’17 Dec ’17 20 20 40 60 80 100 ’17 Mar ’16 Dec ’16 Sep ’16 Jun ’16 Mar 100 80 60 40 20 100 80 60 40 20 100 80 40 60 80 100 60 40 20 100 20 Providers: Transit Zayo NTT Tata Google Netflix Providers: Content Telia XO Vodafone Level3

AT&T - TATA congestion increased and subsided over time

AT&T

Transit Content ’17 Dec ’17 Sep ’17 Jun ’17 Mar ’16 Dec ’16 Sep ’16 Jun ’16 Mar 100 80 60 40 20 100 80 60 40 20

Percent of congested day-links over time

slide-77
SLIDE 77

w w w . cai da.

  • r

38

Content Providers: Content Content Transit Transit Transit Zayo NTT Tata Transit Google Netflix Providers: Content Telia XO Vodafone Level3 Transit

Cox Verizon CenturyLink AT&T Comcast TWC

Content Transit Transit Content Content 80 100 80 60 40 60 40 100 80 60 40 20 ’17 Dec ’17 Sep ’17 Jun 20 100 80 60 40 20 20 100 80 60 40 20 100 80 60 40 20 100 80 60 40 Mar ’16 Jun ’16 Sep ’16 Dec ’16 Mar ’17 Jun ’17 Sep ’17 Dec ’17 20 20 40 60 80 100 ’17 Mar ’16 Dec ’16 Sep ’16 Jun ’16 Mar 100 80 60 40 20 100 80 60 40 20 100 80 40 60 80 100 60 40 20 100 20

Interdomain congestion evolves over time. Need for ongoing measurements!

slide-78
SLIDE 78

w w w . cai da.

  • r

Public Access to Data

  • We are publicly releasing our data via an interactive

visualization system (based on Grafana)

  • And API access to the time series data (based on

InfluxDB)

39

slide-79
SLIDE 79

w w w . cai da.

  • r

Interactive Visualization

40

Longitudinal view of a single link, April - November 2017

slide-80
SLIDE 80

w w w . cai da.

  • r

41

Zooming in for more detail

Interactive Visualization

slide-81
SLIDE 81

w w w . cai da.

  • r

41

Zooming in for more detail

Interactive Visualization

slide-82
SLIDE 82

w w w . cai da.

  • r

Takeaways

  • We have developed a lightweight method and system to

provide third-party visibility into interdomain congestion

  • We hope that our data can provide empirical grounding to

debates over interconnection performance

  • Contact us for access to the data: manic-info@caida.org

42

slide-83
SLIDE 83

w w w . cai da.

  • r

Host a Measurement VP!

We are always looking for volunteers to host VPs!

Contact us:

manic-info@caida.org

43

slide-84
SLIDE 84

w w w . cai da.

  • r

Thanks! Questions? manic-info@caida.org

44