IP Reachability Differences: Myths and Reali<es Dr. - - PowerPoint PPT Presentation

ip reachability differences myths and reali es
SMART_READER_LITE
LIVE PREVIEW

IP Reachability Differences: Myths and Reali<es Dr. - - PowerPoint PPT Presentation

IP Reachability Differences: Myths and Reali<es Dr. Dan Massey Colorado State University He Yan, Ben Say, Dave Oko, Brendan Sheridan Colorado State University Dr. Christos Papadopoulos Dr. Dan Pei


slide-1
SLIDE 1

04/15/2011 ¡ IP ¡Reachability: ¡ ¡Myths ¡and ¡Reality ¡ 1

IP ¡Reachability ¡Differences: ¡ ¡ Myths ¡and ¡Reali<es ¡

  • Dr. Dan Massey

Colorado State University He Yan, Ben Say, Dave Oko, Brendan Sheridan Colorado State University

  • Dr. Christos Papadopoulos
  • Dr. Dan Pei

Colorado State University AT&T Research

slide-2
SLIDE 2

Ideally, ¡all ¡ISPs ¡provide ¡you ¡with ¡ reachability ¡to ¡same ¡Internet ¡

Bogon filtering matters, but the fact is that regional differences matter most and these differences may be political, economic, technical, or a combination of the above In practice: Different ISPs can reach different portions of the Internet. In other words, there are some prefixes that can be reached via ISP A but cannot be reached by ISP B (a vice versa). The reasons why this occur are well known: Bogon filtering causes reachability differences for newly allocated prefixes.

04/15/2011 ¡ 2 IP ¡Reachability: ¡ ¡Myths ¡and ¡Reality ¡

Bogon filtering and regional issues matter, but the real difference is due to various customer only services. Even if there are differences, there are very little (if any) “production” services

  • n networks that are not globally reachable from most ISPs.
slide-3
SLIDE 3

Our ¡Objec<ves ¡

  • 1. Demonstrate ¡reachability ¡differences ¡exist ¡

and ¡quan<fy ¡the ¡extent ¡of ¡these ¡differences. ¡

  • 2. Examine ¡which ¡conjectures ¡(if ¡any) ¡provide ¡a ¡

significant ¡contribu<on ¡to ¡the ¡differences. ¡

04/15/2011 ¡ 3 IP ¡Reachability: ¡ ¡Myths ¡and ¡Reality ¡

slide-4
SLIDE 4

¡Methodology ¡

  • Compare ¡BGP ¡Rou<ng ¡Tables ¡from ¡25 ¡ISPs ¡

– Data ¡collected ¡from ¡Oregon ¡RouteViews ¡ – All ¡peers ¡send ¡full ¡tables ¡of ¡over ¡310K ¡prefixes ¡ – Studied ¡from ¡April ¡1 ¡to ¡April ¡20, ¡2010 ¡

  • Study ¡Uses ¡ISP0 ¡As ¡a ¡Central ¡Point ¡of ¡View ¡ ¡

– ISP0 ¡is ¡a ¡<er-­‑1 ¡ISP ¡ – Have ¡addi<onal ¡data ¡on ¡ISP0 ¡opera<ons ¡ – Compare ¡each ¡of ¡the ¡other ¡24 ¡ISPs ¡to ¡ISP0 ¡

04/15/2011 ¡ 4 IP ¡Reachability: ¡ ¡Myths ¡and ¡Reality ¡

slide-5
SLIDE 5

Types ¡of ¡Table ¡Differences ¡

  • Exact ¡Matches: ¡ ¡ISP0 ¡prefix ¡appears ¡ISPX ¡table ¡ ¡

– Ex: ¡ ¡Both ¡ISP0 ¡and ¡ISP2 ¡have ¡a ¡route ¡for ¡129.82/16 ¡

  • Covered ¡Prefixes: ¡ ¡ISP0 ¡prefix ¡does ¡not ¡appear ¡in ¡

ISPX, ¡ ¡but ¡ISPX ¡can ¡reach ¡all ¡IP ¡addresses ¡

– Less ¡Specific: ¡ ¡131.179.0/17 ¡does ¡not ¡appear ¡in ¡ISPX, ¡ ¡but ¡ 131.179/16 ¡does ¡appear ¡in ¡ISPX ¡ – More ¡specifics: ¡ ¡131.179.0/17 ¡does ¡not ¡appear ¡in ¡ISPX, ¡ ¡ ¡ but ¡131.179.0/18 ¡and ¡131.179.64/18 ¡appear ¡in ¡ISP ¡X ¡

  • Remainder ¡represents ¡poten<al ¡reachability ¡
  • differences. ¡

04/15/2011 ¡ 5 IP ¡Reachability: ¡ ¡Myths ¡and ¡Reality ¡

slide-6
SLIDE 6

Exact ¡Matches ¡

6 IP ¡Reachability: ¡ ¡Myths ¡and ¡Reality ¡ 04/15/2011 ¡

306000 307000 308000 309000 310000 311000 312000 313000 5 10 15 20 25 Number of Prefixes ISP Index

98.5% of ISP0 prefixes also appear in ISPX (not shown here, but 97.25% of ISPX prefixes also appear in ISP0)

slide-7
SLIDE 7

The ¡BGP ¡Table ¡Differences ¡

7 IP ¡Reachability: ¡ ¡Myths ¡and ¡Reality ¡ 04/15/2011 ¡

500 1000 1500 2000 2500 3000 3500 4000 4500 5 10 15 20 25 Number of Prefixes ISP Index

On average, 3250 prefixes appeared in ISP0 tables, but did not appear in ISP22 tables

slide-8
SLIDE 8

Covered ¡Prefixes ¡

8 IP ¡Reachability: ¡ ¡Myths ¡and ¡Reality ¡ 04/15/2011 ¡

1,000 2,000 3,000 4,000 5,000 6,000 7,000 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 Number of Prefixes ISP Index

Not Covered Covered

Nearly all ISP0 prefixes that don’t appear in ISPX are covered by other table entries in ISPX

slide-9
SLIDE 9

Overall ¡Reachability ¡Differences ¡ (Prefixes) ¡

9 IP ¡Reachability: ¡ ¡Myths ¡and ¡Reality ¡ 04/15/2011 ¡

1 10 100 1,000 10,000 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 Number of Prefixes ISP Index

Not Reachable Partially Reachable

slide-10
SLIDE 10

Overall ¡Reachability ¡Differences ¡ (IP ¡Addresses) ¡

10 IP ¡Reachability: ¡ ¡Myths ¡and ¡Reality ¡ 04/15/2011 ¡

1 10 100 1,000 10,000 100,000 1e+06 1e+07 1e+08 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 Number of Destinations (IPs) ISP Index

slide-11
SLIDE 11

Our ¡Objec<ves ¡

  • 1. Reachability ¡differences ¡do ¡exist. ¡

– Small ¡percentage ¡(less ¡than ¡3%) ¡of ¡total ¡routes. ¡ – Can ¡be ¡up ¡to ¡1000 ¡prefixes ¡and ¡millions ¡of ¡IP ¡ addresses ¡ ¡

  • 2. Examine ¡which ¡conjectures ¡(if ¡any) ¡provide ¡a ¡

significant ¡contribu<on ¡to ¡the ¡differences. ¡

04/15/2011 ¡ 11 IP ¡Reachability: ¡ ¡Myths ¡and ¡Reality ¡

slide-12
SLIDE 12

Transient ¡Changes ¡

  • Short ¡Lived ¡BGP ¡Table ¡Differences ¡

– Fast ¡learners: ¡ ¡ ¡IPS0 ¡learns ¡the ¡route ¡first, ¡ ¡ ¡eventually ¡other ¡ ISPs ¡catch ¡up ¡and ¡also ¡learn ¡the ¡route ¡ – Slow ¡to ¡react: ¡ ¡ISPX ¡learned ¡the ¡prefix ¡is ¡unreachable, ¡ ¡but ¡ it ¡takes ¡longer ¡for ¡ISP0 ¡to ¡withdraw ¡the ¡route. ¡

  • Long ¡Term ¡BGP ¡Table ¡Differences. ¡

– For ¡mul<ple ¡days, ¡ ¡ ¡ISP0 ¡could ¡reach ¡the ¡prefix ¡but ¡ISPX ¡ could ¡not. ¡

04/15/2011 ¡ 12 IP ¡Reachability: ¡ ¡Myths ¡and ¡Reality ¡

slide-13
SLIDE 13

Transient ¡Behaviors ¡

13 IP ¡Reachability: ¡ ¡Myths ¡and ¡Reality ¡ 04/15/2011 ¡

500 1,000 1,500 2,000 2,500 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 Number of Prefixes Number of Days Appeared

1/3 of reachability differences are short lived 1/3 of reachability differences persisted for entire study period.

slide-14
SLIDE 14

Bogon ¡Routes ¡

  • ISPs ¡Apply ¡Policies ¡to ¡Filter ¡Invalid ¡Routes ¡

– No ¡universal ¡agreement ¡on ¡what ¡is ¡“invalid” ¡

  • New ¡Alloca<ons ¡and ¡Bogons ¡

– Filter ¡unallocated ¡prefixes ¡and ¡update ¡filter ¡ager ¡alloca<on ¡ – Going ¡back ¡7 ¡months ¡accounts ¡for ¡43 ¡prefixes. ¡ – Going ¡back ¡years ¡yields ¡295 ¡prefixes. ¡

  • Management ¡Strategies ¡and ¡RADb ¡

– 92 ¡prefixes ¡with ¡no ¡entry ¡in ¡RADb ¡ – 199 ¡prefixes ¡with ¡no ¡origin ¡in ¡RADb. ¡

  • Does ¡produce ¡reachability ¡differences, ¡ ¡but ¡only ¡

explains ¡a ¡small ¡percent ¡of ¡differences. ¡

04/15/2011 ¡ 14 IP ¡Reachability: ¡ ¡Myths ¡and ¡Reality ¡

slide-15
SLIDE 15

Regional ¡Differences ¡

15 IP ¡Reachability: ¡ ¡Myths ¡and ¡Reality ¡ 04/15/2011 ¡

500 1,000 1,500 2,000 2,500 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 Number of Prefixes Number of Days Appeared

1895 prefixes from India, only unreachable from ISP 22. ISP 22 added these routes on day 13 of study. Bulgaria accounted for most prefixes (509) Great Britain was second highest (168)

slide-16
SLIDE 16

Customer ¡Services ¡and ¡Prefix ¡Usage ¡

  • ISPs ¡Provide ¡Specialized ¡Services ¡to ¡Customers ¡ ¡

– Confirmed ¡ISP0 ¡has ¡several ¡such ¡prefixes ¡ – Most ¡are ¡covered ¡prefixes, ¡ ¡but ¡did ¡find ¡some ¡uncovered ¡prefixes ¡ – Accounts ¡for ¡9 ¡prefixes ¡in ¡our ¡study. ¡

  • Are ¡Services ¡Running ¡on ¡the ¡Impacted ¡Prefixes? ¡

– Scanned ¡prefixes ¡looking ¡for ¡common ¡services ¡ – Limited ¡scan ¡to ¡/19 ¡or ¡more ¡specifics ¡ – Responding ¡prefixes ¡further ¡checked ¡for ¡DNS ¡entries ¡

  • Contacted ¡Prefix ¡Owners ¡

– Responses ¡include ¡reports ¡of ¡reachability ¡problems, ¡ ¡ ¡blacklist ¡issues, ¡ ¡ network ¡tes<ng, ¡ ¡and ¡so ¡forth ¡

  • Valid ¡Global ¡Services ¡Running ¡on ¡Clear ¡Majority ¡of ¡Prefixes ¡

04/15/2011 ¡ 16 IP ¡Reachability: ¡ ¡Myths ¡and ¡Reality ¡

slide-17
SLIDE 17

RouteViews ¡Data ¡in ¡Real-­‑Time ¡

17 IP ¡Reachability: ¡ ¡Myths ¡and ¡Reality ¡ 04/15/2011 ¡

slide-18
SLIDE 18

Conclusions ¡

  • Reachability ¡Issues ¡Do ¡Exist ¡

– Over ¡97% ¡of ¡prefixes ¡appeared ¡as ¡exact ¡matches ¡ – But ¡differences ¡do ¡occur ¡and ¡impact ¡both ¡research ¡ and ¡opera<ons ¡

  • Common ¡Conjectures: ¡

– Transient ¡issues, ¡ ¡Bogons, ¡ ¡ ¡Management, ¡ ¡ ¡ Regional ¡Policies, ¡ ¡ ¡Configura<on ¡Errors, ¡ ¡Special ¡ Services ¡all ¡proposed ¡as ¡explaina<ons ¡ – All ¡are ¡valid, ¡ ¡but ¡none ¡are ¡dominant. ¡

04/15/2011 ¡ 18 IP ¡Reachability: ¡ ¡Myths ¡and ¡Reality ¡

slide-19
SLIDE 19

Ques<ons ¡

massey@cs.colostate.edu ¡

04/15/2011 ¡ 19 IP ¡Reachability: ¡ ¡Myths ¡and ¡Reality ¡