dig into mpls transit tunnel diversity
play

Dig into MPLS: Transit Tunnel Diversity Yves Vanaubel, Pascal - PowerPoint PPT Presentation

Dig into MPLS: Transit Tunnel Diversity Yves Vanaubel, Pascal Mrindol, Benoit Donnet , Jean-Jacques Pansiot Plane 1 Agenda Motivations Measuring MPLS LPR Evaluation Conclusion AIMS - CAIDA/UCSD - March/April 2015 - Benoit


  1. Dig into MPLS: Transit Tunnel Diversity Yves Vanaubel, Pascal Mérindol, Benoit Donnet , Jean-Jacques Pansiot Plane 1

  2. Agenda • Motivations • Measuring MPLS • LPR • Evaluation • Conclusion AIMS - CAIDA/UCSD - March/April 2015 - Benoit Donnet 2

  3. Motivations • Current studies are mainly about MPLS discovery and its impact on topology discovery B. Donnet, M. Luckie, P. Mérindol, J.-J. Pansiot. Revealing MPLS - Tunnels Obscured from Traceroute . In ACM SIGCOMM Computer Communication Review, 42(2)., pp. 87-93. April 2012 J. Sommers, B. Eriksson, P. Barford. On the Prevalence and - Characteristics of MPLS Deployments in the Open Internet . In Proc. Internet Measurement Conference (IMC). October 2008. T. Flach, E. Katz-Basset, R. Govindan. Quantifying Violations of - Destination-Based Forwarding on the Internet . Proc. Internet Measurement Conference (IMC). November 2012. • Actual usage of MPLS by operators not yet studied AIMS - CAIDA/UCSD - March/April 2015 - Benoit Donnet 3

  4. Motivations (2) • Observation several LSPs may exist for a given <Ingress LER, Egress - LER> LSP 1 : LSP 2 : LSP 1 : Ingress Egress LER LER B D A F C E AIMS - CAIDA/UCSD - March/April 2015 - Benoit Donnet 4

  5. Motivations (3) • Those LSPs between <Ingress LER, Egress LER> can represent load balancing (i.e., ECMP) between the ingress and the - egress LER traffic-engineering - • We want to distinguish 3 types of MPLS tunnels mono-path - transit tunnel without TE - transit tunnel with TE - • We focus on explicit MPLS tunnels - transit tunnels - inter-domain tunnels seems negligible ✓ AIMS - CAIDA/UCSD - March/April 2015 - Benoit Donnet 5

  6. Measuring MPLS • The discovery of MPLS can be based on standard active measurement tools ( [CCR2012] ) ping - traceroute - • Two options are required 1. ICMP extension ( [RFC4950] ) if an MPLS router must forge an ICMP time_exceeded ✓ message, it should quote the MPLS LSE stack in it 2. TTL propagate ( [RFC3443] ) the ingress LER of an MPLS should initialize the LSE-TTL ✓ with the value inside the IP-TTL field (iTTL) the opposite operation is done by the egress LER (oTTL) ✓ AIMS - CAIDA/UCSD - March/April 2015 - Benoit Donnet 6

  7. Measuring MPLS (2) • MPLS Explicit tunnels RFC4950 ∧ RFC3443 - R 1 R 2 R 3 R 4 R 5 Source Ingress Egress Destination LSP PHP LER LER Traceroute output: 1. R 1 2. R 2 - MPLS tag 3. R 3 - MPLS tag 4. R 4 - MPLS tag 5. R 5 6. Destination AIMS - CAIDA/UCSD - March/April 2015 - Benoit Donnet 7

  8. LPR • L abel P attern R ecognition algorithm • Allows to distinguish multi-FEC from IP load balancing • Passive classification method works offline, once the data has been collected - requires no additional probing than traceroute - • Recognizes behaviors of LDP vs. RSVP-TE based on MPLS labels distribution and IP addresses • LPR provides four classes AIMS - CAIDA/UCSD - March/April 2015 - Benoit Donnet 8

  9. LPR (2) • Class 1 MonoLSP - Ingress Egress PHP LER LER L 1 L 2 L 2 L 1 A B C D Trace LSP 1 : Trace LSP 2 : 1. A 1. A 2. B - Label L 1 2. B - Label L 1 3. C - Label L 2 3. C - Label L 2 4. D 4. D Same IP addresses and same Labels AIMS - CAIDA/UCSD - March/April 2015 - Benoit Donnet 9

  10. LPR (3) • Class 2 Multi-FEC - Common IP means an interface Ingress Egress LER PHP LER L 1 B 1 1 2 2 A D E F G L 2 C Trace LSP 1 : Trace LSP 2 : 1. A 1. A 2. … 2. … 3. C - Label 3. B - Label Different labels 4. D 2 - Label 4. D 1 - Label for at least 5. E - Label L 2 5. E - Label L 1 6. … 6. … 1 common IP 7. F 2 - Label 7. F 1 - Label 8. G 8. G AIMS - CAIDA/UCSD - March/April 2015 - Benoit Donnet 10

  11. LPR (4) • Class 3 ECMP Mono-FEC : disjoint routers - Common IP Ingress Egress LER PHP L 2 LER B 1 1 2 2 A D E F G L 2 C Trace LSP 1 : Trace LSP 2 : 1. A 1. A 2. … 2. … 3. C - Label 3. B - Label 4. D 2 - Label L 1 Same label 4. D 1 - Label L 1 5. E - Label L 2 5. E - Label L 2 ∀ common IPs 6. … 6. … 7. F 2 - Label 7. F 1 - Label 8. G 8. G AIMS - CAIDA/UCSD - March/April 2015 - Benoit Donnet 11

  12. LPR (5) • Class 3 (cont.) ECMP Mono-FEC : parallel links - Same labels along all the LSPs Ingress Egress LER L 3 PHP L1 L 2 L 4 LER 1 1 1 2 2 2 A C B F D E L1 L 4 L 2 L 3 Different IPs are aliases! Trace LSP 1 : Trace LSP 2 : 1. A 1. A 2. … 2. … 3. B - Label L 1 3. B - Label L 1 4. C 2 - Label L 2 4. C 1 - Label L 2 5. D 2 - Label L 3 5. D 1 - Label L 3 6. … 6. … 7. E 2 - Label L 4 7. E 1 - Label L 4 8. F 8. F AIMS - CAIDA/UCSD - March/April 2015 - Benoit Donnet 12

  13. LPR (6) • Class 4 Unclassified - • If PHP is used, the Egress LER does not exhibit labels • It may happen that LSPs do not intersect on a common IP address • Those tunnels are arbitrarily tagged as unclassified AIMS - CAIDA/UCSD - March/April 2015 - Benoit Donnet 13

  14. Evaluation • Archipelago platform • First traceroute cycle of each month since 2010 until December 2014 • 60 cycles AIMS - CAIDA/UCSD - March/April 2015 - Benoit Donnet 14

  15. Evaluation (2) • Numbers of tunnels unexplained drop AIMS - CAIDA/UCSD - March/April 2015 - Benoit Donnet 15

  16. Evaluation (3) • AS6453 AIMS - CAIDA/UCSD - March/April 2015 - Benoit Donnet 16

  17. Evaluation (4) • AS1273 AIMS - CAIDA/UCSD - March/April 2015 - Benoit Donnet 17

  18. Evaluation (5) • AS2914 AIMS - CAIDA/UCSD - March/April 2015 - Benoit Donnet 18

  19. Evaluation (6) • AS286 AIMS - CAIDA/UCSD - March/April 2015 - Benoit Donnet 19

  20. Evaluation (7) • Tunnel length AIMS - CAIDA/UCSD - March/April 2015 - Benoit Donnet 20

  21. Evaluation (8) • Tunnel width AIMS - CAIDA/UCSD - March/April 2015 - Benoit Donnet 21

  22. Evaluation (9) • Tunnel symmetry AIMS - CAIDA/UCSD - March/April 2015 - Benoit Donnet 22

  23. Conclusion • New algorithm to reveal TE usage within ASes label distribution - Mono- or Multi-FEC ✓ ECMP load balancing - parallel links or disjoint routers ✓ dynamics - temporal evolution of MPLS deployment and usage ✓ AIMS - CAIDA/UCSD - March/April 2015 - Benoit Donnet 23

  24. Conclusion (2) • Next steps deeper investigation of Multi-FEC class - high frequency traceroute to observe labels behavior ✓ deeper investigation of ECMP class - Paris Traceroute mda mode ✓ providing a library to automatically export MPLS tunnels - usage from CAIDA dataset AIMS - CAIDA/UCSD - March/April 2015 - Benoit Donnet 24

Download Presentation
Download Policy: The content available on the website is offered to you 'AS IS' for your personal information and use only. It cannot be commercialized, licensed, or distributed on other websites without prior consent from the author. To download a presentation, simply click this link. If you encounter any difficulties during the download process, it's possible that the publisher has removed the file from their server.

Recommend


More recommend