OpenIPmap Geolocating Internet Infra-Structure with Inference - - PowerPoint PPT Presentation

openipmap geolocating internet infra structure with
SMART_READER_LITE
LIVE PREVIEW

OpenIPmap Geolocating Internet Infra-Structure with Inference - - PowerPoint PPT Presentation

OpenIPmap Geolocating Internet Infra-Structure with Inference Engines and Crowdsourcing Jasper den Hertog Research and Development RIPE NCC 9 March 2018 | DKNOG8 | Kbenhavn But now for something completely different first Sketches of the


slide-1
SLIDE 1

9 March 2018 | DKNOG8 | København

Jasper den Hertog Research and Development RIPE NCC

OpenIPmap Geolocating Internet Infra-Structure with Inference Engines and Crowdsourcing

slide-2
SLIDE 2

But now for something completely different first

slide-3
SLIDE 3

Jasper den Hertog | 9 March 2018 | DKNOG8

CARNet-AS METRONET-AS ASN-DCM OT-AS T-HT AMISHR DCM-AS SWIPNET VIPNET-AS ASN-ISKON PORTUS-AS DTAG CIX TELEKOM-AT HURRICANE

Sketches of the Peer-to-Peer Fabric

  • f a Country

The quality of end-user connections are often expressed in download speeds towards content providers. Instead, the sketches presented here focus on peer-to-peer connec- tions in a country. The sketches explore the different ways in which end-users are interconnected within the same country: the peer-to-peer fabric. Each sketch rep- resents a snapshot of this fabric at a single given point in time. They try to put a number on the amount of different ways the networks intercon- nect their users. These sketches are created with active measure- ments from the RIPE Atlas measurement plat- form, datasets from RIPEstat, AS-to-ORG datasets from CAIDA and a dataset from APNIC that esti- mates the percentage of end-users in each net- work.

slide-4
SLIDE 4

Jasper den Hertog | 9 March 2018 | DKNOG8

Each network that provides connectivity to more than 1%

  • f the end-users is represented by a colored circle seg-
  • ment. The length of the arc of the segment represents the

percentage of the end-users in a country. The darker green denotes an network for which we have peer-to-peer data. The lighter green color denotes net- works for which we don’t have peer-to-peer data. The open part of the circle rep- resents the sum of all ASes that provide connectivity to less than 1%

  • f the end-users in a country.

Sketches of the Peer-to-Peer Fabric

  • f a Country

The full circle represents 100% of the end-users in a country.

slide-5
SLIDE 5

Jasper den Hertog | 9 March 2018 | DKNOG8

Each ring or circle represents the percentage of the peer-to-peer fabric in a country that passes through this point. The color of the circle or ring denotes the type of location. A green circle on the outer ring represents a network that (mainly) serves end-users. A blue circle on the outer ring represents a network that both serves end-users and provides transit to others end-user net- works within the country. An orange circle in the interior indicates an IXP identi- fied with this country. A blue circle in the interior indi- cates a transit network or an IXP that is external to this country.

Sketches of the Peer-to-Peer Fabric

  • f a Country
slide-6
SLIDE 6

Jasper den Hertog | 9 March 2018 | DKNOG8

Orange lines indicate that two end-user networks are connected through an IXP. Green lines indicate that two end-user networks are directly connected. Blue lines indicate two end-user networks are connected through a transit network. Dotted lines of any color indicate that we cannot fully map this path.

Sketches of the Peer-to-Peer Fabric

  • f a Country
slide-7
SLIDE 7

Jasper den Hertog | 9 March 2018 | DKNOG8

TELIANET-DENMARK DK-ESS-AS STOFANET TDC FIBIA-P-S DK-DANSKKABELTV GlobalConnect-Network PARKNET-AS ASEMNET Hiper Fullrate NIANET-AS TELENOR dk-aura-as hi3g BBN LEVEL3 TELENOR-NEXTEL DIX-DIX LAN TELIANET

Other smaller ASes

Peer-to-Peer Fabric

A network that serves end-users A network that serves end-users and provides transit to other end-user net- works within the country A transit network or an IXP external to this country An IXP that is identified with this country A sizable end-user network for which we have no data A sizable end-user network for which we have data snapshot date

1 March 2018

country

Denmark

github

https://github.com/emileaben/ixp-country-jedi/

url

http://sg-pub.ripe.net/ixp-country-jedi/dk/2018/03/01

slide-8
SLIDE 8

Jasper den Hertog | 9 March 2018 | DKNOG8

POWERVIS-AS-KR SKB-AS LGDACOM KIXS-AS-KR CNM-AS-KR SKTELECOM-NET-AS LGTELECOM-AS-KR A network that serves end-users A network that serves end-users and provides transit to other end-user net- works within the country A transit network or an IXP external to this country An IXP that is identified with this country A sizable end-user network for which we have no data A sizable end-user network for which we have data

Peer-to-Peer Fabric

snapshot date

1 March 2018

country

Southern Korea

github

https://github.com/emileaben/ixp-country-jedi/

url

http://sg-pub.ripe.net/ixp-country-jedi/dk/2018/03/01

slide-9
SLIDE 9

Jasper den Hertog | 9 March 2018 | DKNOG8

EIRCOM AS-BTIRE IBIS-AS LGI-UPC Vodafone-Ireland-ASN DIGIWEB-AS METEORMOBILECOMMS-AS SKYTEL-AS H3GIE TELIANET INEX LAN2 INEX LAN1 CW A network that serves end-users A network that serves end-users and provides transit to other end-user net- works within the country A transit network or an IXP external to this country An IXP that is identified with this country A sizable end-user network for which we have no data A sizable end-user network for which we have data

Peer-to-Peer Fabric

snapshot date

1 March 2018

country

Ireland

github

https://github.com/emileaben/ixp-country-jedi/

url

http://sg-pub.ripe.net/ixp-country-jedi/dk/2018/03/01

slide-10
SLIDE 10

Jasper den Hertog | 9 March 2018 | DKNOG8

ASN-CXA-ALL-CCI-22773-RDC COMCAST-7922 BHN-TAMPA UUNET SCRR-10796 CHARTER-NET-HKY-NC FRONTIER-FRTR ATT-MOBILITY-LLC-AS20057 SCRR-11427 SCRR-11426 SUDDENLINK ROADRUNNER-WEST CENTURYLINK-US-LEGACY-QWEST ATT-INTERNET4 CABLE-NET-1 RR-NYSREGION-ASN-01 SPCS T-MOBILE-AS21928 CELLCO TELIANET NYIIX TWCABLE-BACKBONE NTT-COMMUNICATIONS-2914 Equinix Ashburn AMAZON-02 LEVEL3 SIX Seattle HURRICANE COGENT-174 Equinix New York CHARTER-16787 Equinix San Jose ZAYO-6461 AS6453 XO-AS15 CELLCO-PART ATT-MOBILITY-LLC-AS16572 Equinix Seattle WINDSTREAM GTT-BACKBONE

A network that serves end-users A network that serves end-users and provides transit to other end-user net- works within the country A transit network or an IXP external to this country An IXP that is identified with this country A sizable end-user network for which we have no data A sizable end-user network for which we have data

Peer-to-Peer Fabric

snapshot date

1 March 2018

country

USA

github

https://github.com/emileaben/ixp-country-jedi/

url

http://sg-pub.ripe.net/ixp-country-jedi/dk/2018/03/01

slide-11
SLIDE 11

Why Geolocation of infra-structure?

slide-12
SLIDE 12

Jasper den Hertog | 9 March 2018 | DKNOG8

Why?

Analyse & Optimise Paths

slide-13
SLIDE 13

Jasper den Hertog | 9 March 2018 | DKNOG8

Analyse Paths

slide-14
SLIDE 14

Jasper den Hertog | 9 March 2018 | DKNOG8

Why?

Coherence of Geo Information

slide-15
SLIDE 15

Jasper den Hertog | 9 March 2018 | DKNOG8

Geolocation in the RIPE Database

“Please be aware that this information may not be picked up by geolocation providers.” … “Please be aware that geolocation information is added by the resource holders in the RIPE Database and the RIPE NCC does not verify this information.”

slide-16
SLIDE 16

The problem of Geolocation

slide-17
SLIDE 17

Geolocation is hard

slide-18
SLIDE 18

Jasper den Hertog | 9 March 2018 | DKNOG8

Different Research Approaches

  • Triangulation a.k.a. trilateration
  • Reverse DNS based location inference
  • ‘Administrative’ analyses
  • Verification/falsification procedures
slide-19
SLIDE 19

Jasper den Hertog | 9 March 2018 | DKNOG8

Commercial Offerings

  • Tend to concentrate on end user IP Addresses
  • Opaque Methodology
  • IPv6 address space largely ignored
slide-20
SLIDE 20

Our Integration Attempt

inference engines and crowdsourcing

slide-21
SLIDE 21

Accumulate Research Efforts as Inference Engines

slide-22
SLIDE 22

Jasper den Hertog | 9 March 2018 | DKNOG8

Inference Engines

  • Each of them is applicable only in some cases
  • Each of them has a score factor
  • Complete transparency about the inference methodology
slide-23
SLIDE 23

Jasper den Hertog | 9 March 2018 | DKNOG8

Inference Engines

slide-24
SLIDE 24

Jasper den Hertog | 9 March 2018 | DKNOG8

geolocation API

/locate /worlds /crowdsource /triangulation /anycast /peeringdb

https://openipmap.ripe.net

slide-25
SLIDE 25

Jasper den Hertog | 9 March 2018 | DKNOG8

  • penipmap.ripe.net/api/locate/

83.163.50.165/best

slide-26
SLIDE 26

Jasper den Hertog | 8 March 2018 | DKNOG8

  • penipmap.ripe.net/api/locate/

83.163.50.165/partials

slide-27
SLIDE 27

Jasper den Hertog | 9 March 2018 | DKNOG8

/locate - Active geolocation

If the IP has not been measured yet, a new Ping measurement starts

  • Peering DB data and BGP data are used to reduce the

locations probed

  • Score based on RTT, only RTT <10ms are considered
  • PeeringDB facilities and population boost the score
  • A list of possible locations will be returned
  • We are working on it! (Contributions are welcome!)
slide-28
SLIDE 28

Jasper den Hertog | 9 March 2018 | DKNOG8

AS59626 AS16245

let’s geolocate 2a07:1480:2:303::1

First step: get a set of ASes that are related to this IP address

slide-29
SLIDE 29

Jasper den Hertog | 9 March 2018 | DKNOG8

let’s geolocate 2a07:1480:2:303::1

AS59626 AS16245 Second step: get a set of geographic locations related with these ASes

slide-30
SLIDE 30

Jasper den Hertog | 9 March 2018 | DKNOG8

let’s geolocate 2a07:1480:2:303::1

AS59626 AS16245 Third step: run ping measurements on RIPE Atlas probes within these ASes and these locations.

(2.1ms!)

(6,4ms) fromlocation(6,4ms)

slide-31
SLIDE 31

Jasper den Hertog | 9 March 2018 | DKNOG8

let’s geolocate 2a07:1480:2:303::1

slide-32
SLIDE 32

Crowdsourcing with OpenIPmap

slide-33
SLIDE 33

Jasper den Hertog | 9 March 2018 | DKNOG8

Web UI: put stuff on a map

  • Make sense of RTTs in one glimpse
  • Understand network relationships among countries
  • Verify geographical optimisation policies
  • Crowdsource IP addresses to geolocation
slide-34
SLIDE 34

Jasper den Hertog | 9 March 2018 | DKNOG8

…actually we have two interfaces

https://openipmap.ripe.net https://atlas.ripe.net/measurements/ <TRACEROUTEMSM>

slide-35
SLIDE 35

Jasper den Hertog | 9 March 2018 | DKNOG8

The first one we already saw…

slide-36
SLIDE 36

Jasper den Hertog | 9 March 2018 | DKNOG8

https://atlas.ripe.net/measurements/11585456/#!openipmap

The second shows traceroutes

slide-37
SLIDE 37

Jasper den Hertog | 9 March 2018 | DKNOG8

elements

sidebar map

slide-38
SLIDE 38

Jasper den Hertog | 9 March 2018 | DKNOG8

sidebar

hopswithenrichedinfo Geographical information

slide-39
SLIDE 39

Jasper den Hertog | 9 March 2018 | DKNOG8

map

linesconnectinggeolocatedhosts: hoplines geolocatedhost geolocatedcountry

slide-40
SLIDE 40

Jasper den Hertog | 9 March 2018 | DKNOG8

interaction sidebar and map

double-clickhopline showstracerouteand highlightshopsalongthisline

slide-41
SLIDE 41

Jasper den Hertog | 9 March 2018 | DKNOG8

to each traceroute its own line

slide-42
SLIDE 42

Jasper den Hertog | 9 March 2018 | DKNOG8

data doubt: scores for locations

scoreindicatorinsidebar

slide-43
SLIDE 43

Jasper den Hertog | 9 March 2018 | DKNOG8

crowdsource: add locations

clickhereand… starttypingacityor… acountry

slide-44
SLIDE 44

Jasper den Hertog | 9 March 2018 | DKNOG8

confirm and boost location score

boostthescoreforthislocation

slide-45
SLIDE 45

Jasper den Hertog | 9 March 2018 | DKNOG8

Change and improve locations

starttypingand… …chooseanewormore specificlocationand… nowyoucansubmit!

slide-46
SLIDE 46

Jasper den Hertog | 9 March 2018 | DKNOG8

Change and improve

instantchangesinmapand… allotherhopsthathavethishost

slide-47
SLIDE 47

Jasper den Hertog | 9 March 2018 | DKNOG8

Future Work

New Inference Engines

  • Increase research collaborations
  • Integrate RIR data
  • Reverse DNS engine
  • Attempt to separate infra-structure IP addresses from end-

user IP addresses

  • Automated IP address discovery (IPv6!)
slide-48
SLIDE 48

Jasper den Hertog | 9 March 2018 | DKNOG8

https://openipmap.ripe.net

slide-49
SLIDE 49

Questions

jdenhertog@ripe.net

twitter:tracerouteoftheday!

@density215