BGP Path Exploration Damping (PED)

Transcription

1 BGP Path Exploration Damping (PED) Mattia Rossi Centre for Advanced Internet Architectures (CAIA) Swinburne University of Technology Outline Introduction Motivation Path Exploration Damping - PED Experimental results Reduction of update load MRAI and PED convergence time compared Conclusions and future work IETF-81 mrossi@swin.edu.au 28 July,

2 Outline Introduction Motivation Path Exploration Damping - PED Experimental results Reduction of update load MRAI and PED convergence time compared Conclusions and future work IETF-81 mrossi@swin.edu.au 28 July, About this talk Selected for the Applied Networking Research Prize (ANRP) based on a peer-reviewed paper: Paper published in the IEEE Journal on Selected Areas in Communications (JSAC), October A Technique for Reducing BGP Update Announcements through Path Exploration Damping. Geoff Huston, Mattia Rossi, Grenville Armitage Project sponsored by the Cisco University Research Program Fund 1 IETF-81 mrossi@swin.edu.au 28 July,

3 hat is Path Exploration Damping? Path Exploration Damping PED Algorithm intended to replace MRAI and RFD Methods to reduce update churn and convergence time in BGP BGP is the de-facto standard for inter-domain routing BGP Border Gateway Protocol MRAI Minimum Route Advertisment Interval RFD Route Flap Damping IETF July, PED implementations CAIA Implemented in Quagga Patch available for Quagga Download at Cisco (by Mohammed Mirza) Implemented in CISCO IOS-XE Experimental Version 15.1 Running on Cisco ASR1002 (2RU) Currently tested at APNIC Pty. Ltd., Australia IETF-81 mrossi@swin.edu.au 28 July,

4 Outline Introduction Motivation Path Exploration Damping - PED Experimental results Reduction of update load MRAI and PED convergence time compared Conclusions and future work IETF-81 mrossi@swin.edu.au 28 July, Simplified BGP speaker design Router 1 Router 2 Adj-RIB-in RIB Adj-RIB-out Decision process Router 3 Router 4 RIB: Routing Information Base Routing Table Adj-RIB-(in,out): Adjacency RIB (in,out) IETF-81 mrossi@swin.edu.au 28 July,

5 Basic BGP dynamics IETF July, Basic BGP dynamics Path: 1 IETF-81 mrossi@swin.edu.au 28 July,

6 Basic BGP dynamics Path: 2,1 Path: 1 Path: 2,1 IETF-81 mrossi@swin.edu.au 28 July, Basic BGP dynamics Path: *2,1 3,2,1 Path: 1 Path: 2,1 Path: 3,2,1 IETF-81 mrossi@swin.edu.au 28 July,

7 Basic BGP dynamics Path: *2,1 3,2,1 4,3,2,1 Path: 1 Path: 2,1 Path: 3,2,1 IETF-81 mrossi@swin.edu.au 28 July, Path Exploration Path: *2,1 3,2,1 4,3,2,1 Path: 2,1 Path: 3,2,1 IETF-81 mrossi@swin.edu.au 28 July,

8 Path Exploration Path: *3,2,1 4,3,2,1 Path: 3,2,1 IETF July, Path Exploration Path: *4,3,2,1 A A: 5,3,2,1 IETF-81 mrossi@swin.edu.au 28 July,

9 Path Exploration A A: 5,3,2,1 A: 5,4,3,2,1 IETF-81 mrossi@swin.edu.au 28 July, Path Exploration A: 5,3,2,1 A: 5,4,3,2,1 IETF-81 mrossi@swin.edu.au 28 July,

10 hat is Path Exploration? An update sequence lengthening the AS-path gradually until stability is reached IETF July, MRAI and RFD Need to decrease BGP chattiness and Path Exploration Minimum Route Advertisment Interval MRAI (RFC 1771) Apply 30s (default) delay on announcements Router 1 Router 2 Adj-RIB-in RIB Adj-RIB-out Decision process MRAI Timer MRAI Timer Router 3 Router 4 IETF-81 mrossi@swin.edu.au 28 July,

11 Minimum Route Advertisment Interval MRAI IETF July, Minimum Route Advertisment Interval MRAI aiting 30s Path: 1 30 sec. IETF-81 mrossi@swin.edu.au 28 July,

12 Minimum Route Advertisment Interval MRAI Path: 2,1 30 sec. aiting 60s Path: 1 Path: 2,1 IETF-81 mrossi@swin.edu.au 28 July, Minimum Route Advertisment Interval MRAI Path: *2,1 3,2,1 Delay 0-90s Path: 1 30 sec. Path: 2,1 Path: 3,2,1 IETF-81 mrossi@swin.edu.au 28 July,

13 Minimum Route Advertisment Interval MRAI Path: *2,1 3,2,1 4,3,2,1 Delay 0-90s Path: 1 30 sec. Path: 2,1 Path: 3,2,1 IETF-81 mrossi@swin.edu.au 28 July, Minimum Route Advertisment Interval MRAI Path: *2,1 3,2,1 4,3,2,1 Path: 2,1 Path: 3,2,1 IETF-81 mrossi@swin.edu.au 28 July,

14 Minimum Route Advertisment Interval MRAI Path: *3,2,1 4,3,2,1 Path: 3,2,1 IETF July, Minimum Route Advertisment Interval MRAI Path: *4,3,2,1 Update delayed IETF July,

15 Minimum Route Advertisment Interval MRAI Update delayed IETF July, Minimum Route Advertisment Interval MRAI IETF July,

16 Minimum Route Advertisment Interval MRAI Path: *4,3,2,1 Update delayed IETF July, Minimum Route Advertisment Interval MRAI A A: 5,4,3,2,1 IETF-81 mrossi@swin.edu.au 28 July,

17 Minimum Route Advertisment Interval MRAI A: 5,4,3,2,1 IETF July, MRAI and RFD Need to decrease BGP chattiness and Path Exploration Minimum Route Advertisment Interval MRAI (RFC 1771) Apply 30s (default) delay on announcements MRAI on withdrawals (RATE) allowed per RFC 4271 IETF July,

18 MRAI on withdrawals RATE Path: *2,1 3,1 4,3,1 Path: *1 Path: *3,1 IETF July, MRAI on withdrawals RATE Path: *2,1 3,1 4,3,1 30 sec. Path: *1 Path: *3,1 IETF July,

19 MRAI on withdrawals RATE Path: *2,1 3,1 4,3,1 Data loss! Black hole!! Path: *1 Path: *3,1 IETF July, MRAI and RFD Need to decrease BGP chattiness and Path Exploration Minimum Route Advertisment Interval MRAI (RFC 1771) Apply 30s (default) delay on announcements MRAI on withdrawals (RATE) allowed per RFC 4271 Route Flap Damping RFD (RFC 2439) Flapping = sequence of announcements and withdrawals Suppress flapping prefixes for 1 hour (or more) IETF-81 mrossi@swin.edu.au 28 July,

20 Route Flap Damping RFD Router 1 Flapping prefix! Router 2 Adj-RIB-in RIB Adj-RIB-out RFD RFD Decision process Router 3 Router 4 IETF-81 mrossi@swin.edu.au 28 July, Route Flap Damping RFD Path: *3,2,1 Path: 3,2,1 IETF-81 mrossi@swin.edu.au 28 July,

21 Route Flap Damping RFD A A: 5,3,2,1 Path: 3,2,1 IETF-81 mrossi@swin.edu.au 28 July, Route Flap Damping RFD Path: *4,3,2,1 A: 5,3,2,1 IETF-81 mrossi@swin.edu.au 28 July,

22 Route Flap Damping RFD A A: 5,3,2,1 A: 5,4,3,2,1 IETF-81 mrossi@swin.edu.au 28 July, Route Flap Damping RFD A: 5,3,2,1 A: 5,4,3,2,1 IETF-81 mrossi@swin.edu.au 28 July,

23 Route Flap Damping RFD Path: *2,1 3,2,1 4,3,2,1 RFD suppresses /8 for 1 hour Path: 1 Path: 2,1 Path: 3,2,1 here is my data? IETF-81 mrossi@swin.edu.au 28 July, hat are the problems? MRAI exhibits unpredictable behavior RATE creates black holes RFD penalizes prefix owners even if the misbehavior happened further upstream e want something better! IETF-81 mrossi@swin.edu.au 28 July,

24 Outline Introduction Motivation Path Exploration Damping - PED Experimental results Reduction of update load MRAI and PED convergence time compared Conclusions and future work IETF-81 mrossi@swin.edu.au 28 July, PED algorithm Delay BGP announcements if the announced AS path is longer than the previously known AS path IETF-81 mrossi@swin.edu.au 28 July,

25 PED algorithm illustrated Router 1 Router 2 Adj-RIB-in RIB Adj-RIB-out Decision process PEDI Timer PEDI Timer Router 3 Router 4 PEDI: Path Exploration Damping Interval IETF-81 mrossi@swin.edu.au 28 July, PED algorithm illustrated Router 1 Router 2 Adj-RIB-in RIB Adj-RIB-out Decision process per-peer setting per-prefix application per-peer setting per-prefix application Router 3 Router 4 IETF-81 mrossi@swin.edu.au 28 July,

26 PED algorithm illustrated Router 1 A: /8 first announcement / previously withdrawn Router 2 Adj-RIB-in RIB Adj-RIB-out Decision process announce new best path no delay Router 3 Router 4 IETF-81 mrossi@swin.edu.au 28 July, PED algorithm illustrated Router 1 update Router 2 Adj-RIB-in RIB Adj-RIB-out Decision process longer path! queue update, start PEDI timer Router 3 Router 4 IETF-81 mrossi@swin.edu.au 28 July,

27 PED algorithm illustrated Router 1 update Router 2 Adj-RIB-in RIB Adj-RIB-out Decision process shorter path! send update, delete PEDI timer Router 3 Router 4 IETF-81 mrossi@swin.edu.au 28 July, PED algorithm illustrated Router 1 update Router 2 Adj-RIB-in RIB Adj-RIB-out Decision process withdrawal! send update, delete PEDI timer Router 3 Router 4 IETF-81 mrossi@swin.edu.au 28 July,

28 Outline Introduction Motivation Path Exploration Damping - PED Experimental results Reduction of update load MRAI and PED convergence time compared Conclusions and future work IETF-81 mrossi@swin.edu.au 28 July, PED Data analysis Experiments using 24 hours of real BGP updates Two datasets: 1. APNIC Pty. Ltd., Australia (2 peers) 2. University of Oregon Routeviews, US (5 peers) AS 293 AS AS 3727 AS 4777 AS 4608 AS 2914 AS AS 6447 Routeviews APNIC AS IETF-81 mrossi@swin.edu.au 28 July,

29 PED Data analysis Replayed using the Quagga-Accelerator PED Collector BGP cloud Router 1 Router 2 Quagga Accelerator IETF-81 mrossi@swin.edu.au 28 July, Outline Introduction Motivation Path Exploration Damping - PED Experimental results Reduction of update load MRAI and PED convergence time compared Conclusions and future work IETF-81 mrossi@swin.edu.au 28 July,

30 5 s ec PEDI 0 s ec MRAI 75 s ec PEDI 70 s ec PEDI 65 s ec PEDI 60 s ec PEDI 55 s ec PEDI 50 s ec PEDI 45 s ec PEDI 40 s ec PEDI 35 s ec PEDI 30 s ec PEDI 25 s ec PEDI 20 s ec PEDI 15 s ec PEDI 10 s ec PEDI 30 s ec MRAI Reduction of update load APNIC num ber of m es s ages 160k 140k 120k 100k 80k 60k 40k 20k 20% 29% BGP update m es s ages Pr efix announcem ents Pr efix withdr awals 0 70 s ec PEDI 65 s ec PEDI 60 s ec PEDI 55 s ec PEDI 50 s ec PEDI 45 s ec PEDI 40 s ec PEDI 35 s ec PEDI 30 s ec PEDI 0 s ec MRAI 30 s ec MRAI IETF-81 mrossi@swin.edu.au 28 July, Reduction of update load Routeviews num ber of m es s ages 150k 100k 50k 18% BGP update m es s ages Pr efix announcem ents Pr efix withdr awals 32% 0 IETF-81 mrossi@swin.edu.au 28 July,

31 Outline Introduction Motivation Path Exploration Damping - PED Experimental results Reduction of update load MRAI and PED convergence time compared Conclusions and future work IETF-81 mrossi@swin.edu.au 28 July, Convergence time approximation Control Plane convergence (Optimality) Data plane convergence Forwarding path (Reachability) Convergence measured from a single point of view (Optimality) Router 2 Convergence defined as a route being stable for 5 minutes 240s 30s 30s 30s 120s 5min Time to converge IETF-81 mrossi@swin.edu.au 28 July,

32 Optimality approximation APNIC data Upper and lower quartile Mean Median 500 s econds s PEDI 65s PEDI 60s PEDI 55s PEDI 50s PEDI 45s PEDI 40s PEDI 35s PEDI 30s PEDI 0s MRAI 30s MRAI IETF-81 mrossi@swin.edu.au 28 July, Optimality approximation Routeviews data seconds Upper and lower quartile Mean Median sec PDI 65 sec PDI 60 sec PDI 55 sec PDI 50 sec PDI 45 sec PDI 40 sec PDI 35 sec PDI 30 sec PDI 25 sec PDI 20 sec PDI 15 sec PDI 10 sec PDI 5 sec PDI 0 sec MRAI 30 sec MRAI IETF-81 mrossi@swin.edu.au 28 July,

33 Detailed convergence time analysis Investigate Reachability vs. Optimality Analysis of a whole BGP system Impact on convergence of 4 events causing instability: Link failure along the path alternative path exists Link recovery Prefix withdrawal Prefix announcement IETF-81 mrossi@swin.edu.au 28 July, Detailed convergence time analysis Experimental analysis over 20 testruns Simple example topology ASes are single BGP speakers Example: 30s MRAI on all ASes and 35s PEDI on all ASes AS Prefix Origin: / IETF-81 mrossi@swin.edu.au 28 July,

34 Optimality Announcement of initial route Stable System: * 1 * 2,1 *3,2,1 *4,3,2,1 6,13,12,11,10,1 *13,12,11,10,1 5,4,3,2,1 AS 6 Data to * 1 *10,1 *11,10,1 22,21,20,10, Prefix Origin: / *12,11,10,1 *10,1 *20,10,1 *21,20,10,1 12,11,10,1 Announcement of initial route at AS 6 : PED: 0 seconds MRAI: seconds IETF-81 mrossi@swin.edu.au 28 July, Reachability Link failure Link failure between 0 and 1 : * 1 * 2,1 *3,2,1 *4,3,2,1 6,13,12,11,10,1 *13,12,11,10,1 5,4,3,2,1 AS 6 Data to * 1 *12,22,21,20,10,1 *22,21,20,10, Prefix Origin: / *12,11,10,1 *10,1 *20,10,1 *21,20,10,1 Reachability achieved (1 ) PED: 0 seconds MRAI: 0-4 or seconds IETF-81 mrossi@swin.edu.au 28 July,

35 Optimality Link failure Link failure between 0 and 1 : * 1 * 2,1 *3,2,1 *4,3,2,1 *5,4,3,2,1 13,12,22,21,20,10,1 AS 6 Data to * 1 *12,22,21,20,10,1 *22,21,20,10, Prefix Origin: / *12,22,21,20,10,1 *10,1 *20,10,1 *21,20,10,1 Optimality achieved: PED: 66 seconds (+-jitter) MRAI: 2-58 seconds IETF-81 mrossi@swin.edu.au 28 July, Optimality Link recovery Link recovery between 0 and 1 : * 1 * 2,1 *3,2,1 *4,3,2,1 6,13,12,11,10,1 *13,12,11,10,1 5,4,3,2,1 AS 6 Data to * 1 *10,1 *11,10,1 22,21,20,10, Prefix Origin: / *12,11,10,1 *10,1 *20,10,1 *21,20,10,1 12,11,10,1 Optimality achieved: PED: 0 seconds MRAI: and seconds IETF-81 mrossi@swin.edu.au 28 July,

36 Optimality/Reachability Prefix withdrawn Prefix withdrawal at AS Prefix Origin: / Optimality achieved (route withdrawn on every AS): PED: 0 seconds MRAI: 0 seconds IETF-81 mrossi@swin.edu.au 28 July, Optimality/Reachability Prefix (re)-announced Prefix announcement at * 1 * 2,1 *3,2,1 *4,3,2,1 6,13,12,11,10,1 *13,12,11,10,1 5,4,3,2,1 AS 6 Data to * 1 *10,1 *11,10,1 22,21,20,10, Prefix Origin: / *12,11,10,1 *10,1 *20,10,1 *21,20,10,1 12,11,10,1 Optimality achieved (same as initial announcement): PED: 0 seconds MRAI: 32-34, 58-60, seconds IETF-81 mrossi@swin.edu.au 28 July,

37 Outline Introduction Motivation Path Exploration Damping - PED Experimental results Reduction of update load MRAI and PED convergence time compared Conclusions and future work IETF-81 mrossi@swin.edu.au 28 July, Conclusions In this talk: PED decreases update load PED converges to Reachability as fast or faster than MRAI PED converges to Optimality slower than MRAI in one case In the paper: PED interacts well with MRAI PED can be deployed incrementally A single PED speaker is beneficial to the BGP system 35s PEDI is a safe defaulte value in the MRAI dominated Internet In the future: Dynamic PEDI per prefix More heuristics IETF-81 mrossi@swin.edu.au 28 July,