1. Monitoring Persistently Congested Internet Links Leiwen (Karl) Deng Aleksandar Kuzmanovic Northwestern University http://networks.cs.northwestern.edu

2. Karl Deng Monitoring Persistently Congested Internet Links 2 Objective New probing methods that can improve measurement observability for core congestion Pong – a tool specialized in measuring a subset of non-edge links exhibiting repetitive congestion –Can reveal systematic problems such as routing pathologies, poorly-engineered network policies, or non-cooperative inter-AS relationships Lightweight: monitoring in addition to on- demand measuring –A building block of a large scale triggered monitoring system for Internet congestion

3. Karl Deng Monitoring Persistently Congested Internet Links 3 Repetitive Congestion We focus on locating and monitoring non-edge links that exhibit repetitive congestion –Queuing delay as congestion indicator –Queue building-up repetitively happens on time scales of one or more minutes.

4. Karl Deng Monitoring Persistently Congested Internet Links 4 Methodology Highlights Coordinated probing –Probe from both endpoints of a path –Combine end-to-end probes with (TTL limited) probes to intermediate routers Infer underlying path topology conditions –Implicit inference –Based on measured queuing delays on different probing paths Use statistics over longer time scales Quantify measurement accuracy –Link measurability score

5. Karl Deng Monitoring Persistently Congested Internet Links 5 SD Probe f s d b f (“forward”) probeb (“backward”) probe s (“source”) probe d (“destination”) probe,,, A Simplified Case – Symmetric Path Coordinated Probing

6. Karl Deng Monitoring Persistently Congested Internet Links 6 SD f s d b ΔfsΔfs ΔfdΔfd Half-path queuing delay Locating Congested Links Tracing Congestion Status Probe ΔdΔd ΔbΔb ΔfΔf ΔsΔs Coordinated Probing

7. Karl Deng Monitoring Persistently Congested Internet Links 7 Infer Underlying Path Topology Conditions SD f s d b Probe ΔdΔd ΔbΔb ΔfΔf ΔsΔs Condition: Δf +Δb ≈Δs +Δd Path Pattern: 4-p probing scenario

8. Karl Deng Monitoring Persistently Congested Internet Links 8 S D f s b d Observed by b probe only Paired d probe Congestion Pair up Fsd probing S D f s b No suitable d probes to pair up with this s probe Congestion Fsb probing Probing techniqueCondition 4-p probingΔf +Δb ≈Δs +Δd Δf ≈Δs +ΔdFsd probing Δs ≈Δf +ΔbFsb probing unconditional2-p probing S D f s b d Paired d probe Congestion 4-p probing Pair up ΔfsΔfs ΔfdΔfd Estimates of half-path queuing delay Probe Infer Underlying Path Topology Conditions

9. Karl Deng Monitoring Persistently Congested Internet Links 9 Probing technique Condition 4-p probing Fsd probing Fsb probing 2-p probing Δf +Δb ≈Δs +Δd Δf ≈Δs +Δd Δs ≈Δf +Δb unconditional max(Δf +Δb, Δs +Δd) QoM 4p = 1 − |(Δf +Δb) − (Δs +Δd)| max(Δf, Δs +Δd) QoM fsd = 1 − |(Δf − (Δs +Δd)| QoM fsb = 1 − |(Δs − (Δf +Δb)| max(Δs, Δf +Δb) Demote Promote (Last resort) Definition of QoM Select Probing Techniques Adjust probing technique online based on quality of measurability (QoM)

10. Karl Deng Monitoring Persistently Congested Internet Links 10 SD Probe ΔfsΔfs ΔfdΔfd ΔfsΔfs ΔfdΔfd ΔfsΔfs ΔfdΔfd ΔfsΔfs ΔfsΔfs ΔfdΔfd ΔfdΔfd Probe all nodes simultaneously Locating Congested Links Perform coordinated probing for all intermediate nodes

11. Karl Deng Monitoring Persistently Congested Internet Links 11 Correlate probes to neighboring nodes SD Probe Switch Point Approach Congested link is identified Congestion Locating Congested Links

12. Karl Deng Monitoring Persistently Congested Internet Links 12 SD Link C (Identified congested link) Link C Congestion Status Time Congestion Use fast rate end-to-end probing Tracing Congestion Status f f f f f f f f f f f f f f f

13. Karl Deng Monitoring Persistently Congested Internet Links 13 12 3 4 5 6 7 8 1 34 67 2 5 0.37s on/off 0.71s on/off 0.53s on/off 0.47s on/off0.83s on/off 12 1110 9 8 119 Topology: 12 nodes (PCs), 11 links Link: 100Mbps, 2ms Cross traffic: Each consists of 3 parallel TCP flows, 50% time on and 50% time off. Build multiple bottlenecks: Cross traffics are added to corresponding links concurrently. Emulab Experiment Example

14. Karl Deng Monitoring Persistently Congested Internet Links 14 12 3 4 5 6 7 8 1 34 67 2 5 0.37s on/off 0.71s on/off 0.53s on/off 12 1110 9 8 119 At the Beginning Emulab Experiment Example

15. Karl Deng Monitoring Persistently Congested Internet Links 15 12 3 4 5 6 7 8 1 34 67 2 5 0.37s on/off 0.71s on/off 0.53s on/off 0.47s on/off0.83s on/off 12 1110 9 8 119 After adding backward bottlenecks Emulab Experiment Example

16. Karl Deng Monitoring Persistently Congested Internet Links 16 12 3 4 5 6 7 8 1 34 67 2 5 0.37s on/off 0.71s on/off 0.53s on/off 12 1110 9 8 119 0.29s on/off0.63s on/off After adding two more forward bottlenecks Emulab Experiment Example

17. Karl Deng Monitoring Persistently Congested Internet Links 17 Optimizing Pong in the Internet Set queuing delay threshold –Based on distribution of queuing delay samples Tune other parameters based on experiments on the PlanetLab Minimize measurement errors –Detect and react to anomalies (clock skews, router alterations, ICMP queuing, etc) –Use instantaneous quality of measurement value as sample weight Quantify measurement quality –Help select vantage points

18. Karl Deng Monitoring Persistently Congested Internet Links 18 Quantify Measurement Quality Help select vantage points Link measurability score 1.Probing technique and quality of measurability 2.Queuing delay threshold quality 3.Observability score Congestion observed on a less frequently congested link can be blurred by a much more frequently congested link on the same path.

19. Karl Deng Monitoring Persistently Congested Internet Links 19 Conclusion Pong – a tool specialized in measuring a subset of non-edge links exhibiting repetitive congestion Coordinated probing Infer underlying path topology conditions Select probing techniques online –Quality of measurability Quantify measurement quality –Link measurability score

20. Karl Deng Monitoring Persistently Congested Internet Links 20 Thank you! Questions?