empirical analysis of the effects and the mitigation of ...empirical analysis of the effects and the...

Empirical Analysis of the Effects and the Mitigation of IPv4 Address Exhaustion

wissenschaftliche Aussprache2. August 2017

Philipp Richter

The Internet connects 3.5 billion people as of 2016. (48% of world population)

Internet Penetration, 2017, ISOC.

1

●●

●

●

●

●

●

●

●

●

●

●

year

user

s/su

bscr

iber

s [b

illion

s]

2005

2006

2007

2008

2009

2010

2011

2012

2013

2014

2015

2016

01B

2B3B

4B ● Internet usersmobile broadband subscriptionsfixed broadband subscriptions

2

The Internet Protocol Suite

3

The Narrow Waist of the Protocol Stack

‣ Original design: One IP address per host

‣ IPv4: 32-bit addresses, est. 1981~ 4B unique IPv4 addresses

‣ Today: 3.5B users, ~7B connected devices.

3

IPv4 Exhaustion received a lot of attention.But little in terms systematic empirical assessment.

IPv4 Address Exhaustion

4

ACM CCR ’15 (Best of CCR)

Systematic Framing of IPv4 Address Exhaustion


Systematic Framing of IPv4 Address Exhaustion

‣ IPv4 addresses need to be globally unique

‣ We need a management body that distributes them

10

Early Registration Needs-Based Provision Depletion & Exhaustion

• Informal Distribution• Scarcity minor issue• Non-commercial Internet

• Distribution process• Justification of need• ISPs don’t pay for IPs

• 4 out of 5 RIRs depleted• Address Markets• Transfer Policies

1981 ~1995 ~2011A History of IPv4 Address Block Management

● ● ● ● ● ● ● ● ●●

● ● ● ● ● ● ● ● ● ● ● ● ● ● ●●

●●

●● ● ● ● ● ●

allocated IPv4 addresses

1985 1990 1995 2000 2005 2010 2015

routable IPv4 addresses(3.7B)

5

10





1981 ~1995 ~2011A History of IPv4 Address Block Management

● ● ● ● ● ● ● ● ●●

● ● ● ● ● ● ● ● ● ● ● ● ● ● ●●

●●

●● ● ● ● ● ●


1985 1990 1995 2000 2005 2010 2015


40% of the space given outby ~1995

“LEGACY” space

5

10





1981 ~1995 ~2011

● ● ● ● ● ● ● ● ●●

● ● ● ● ● ● ● ● ● ● ● ● ● ● ●●

●●

●● ● ● ● ● ●


1985 1990 1995 2000 2005 2010 2015


A History of IPv4 Address Block Management

5

10


• Informal Distribution • Scarcity minor issue • Non-commercial Internet

• Distribution process • Justification of need • ISPs don’t pay for IPs

• 4 out of 5 RIRs depleted • Address Markets • Transfer Policies

1981 ~1995 ~2011

A History of IPv4 Address Block Management

● ● ● ● ● ● ● ● ●●

● ● ● ● ● ● ● ● ● ● ● ● ● ● ●●

●●

●● ● ● ● ● ●

cumulative yearly allocations

1985 1990 1995 2000 2005 2010 2015


Network operators around the world need to find ways to mitigate their IPv4 scarcity issues.


Use IPv4 space more efficiently

Multiplex IPv4:Carrier-Grade NAT Transition to IPv6

Systematic Framing of IPv4 Exhaustion

mitigation strategies

Strategy (i): Use IPv4 space more efficiently

Hypothesis

‣ IPv4 space not fully utilized‣ Underutilized space could be freed up and used/transferred

Research Questions

‣ What is the potential for utilization increase?‣ Which knobs could be adjusted here?

6

Degrees of Address Use

Allocation~99%registered to a network?

7

Allocation

Routing

~99%

advertised in the global routing table?


registered to a network?

7

IPv4 Address Activity: Global Routing Table/8

equ

ivale

nts

● ●●

●●

●●

●

●

●

●

●

●

●●

●● ● ●

0

50

100

150

200

250

Nov 1997 Jan 2001 Jan 2005 Jan 2009 Jan 2013 Jan 2017

routable address space limit (220.7 /8 equivalents)

total address space limit (256 /8 equivalents)

● allocated address blocksrouted address blocks

8


equ

ivale

nts

● ●●

●●

●●

●

●

●

●

●

●

●●

●● ● ●

0

50

100

150

200

250





Mostly LEGACY (pre-1995) allocations

8


equ

ivale

nts

● ●●

●●

●●

●

●

●

●

●

●

●●

●● ● ●

0

50

100

150

200

250





Mostly LEGACY (pre-1995) allocations

Impact of Internet Governance! 8


Allocation

Routing

~99%

~75%advertised in the global routing table?


9

Allocation

Routing

Activity

~99%


actively used?



9

1101010101111010101011

Passive Measurements Active Measurements

Measuring IPv4 Activity

“How many IPv4 address blocks show activity?”

10

IPv4 Activity - Counting Active Addresses

total active: 36% (5.3M /24s)

Our preliminary study

31% (4.5M /24s)‣ 4 passive vantage points

‣ 3 active scanning campaigns

33% (4.8M /24s)

Overlap, but each vantage point has unique contribution

11

IPv4 Activity - Counting Active Addresses


Our preliminary study

31% (4.5M /24s)‣ 4 passive vantage points

‣ 3 active scanning campaigns

33% (4.8M /24s)

Related Work (Zander et al.)


‣ 7 passive vantage points & 2 active campaigns

11

Overlap, but each vantage point has unique contribution

Allocation

Routing

Activity

~99%


actively used?lower bound:

~36-41% (/24s)



12

Allocation

Routing

Activity

~99%


actively used?lower bound:

~36-41% (/24s)


Significant potential for increasing the utilizationof the IPv4 address space


12

IPv4 Address Activity from a CDN

The CDN Vantage Point

‣ 200,000+ servers in 1500+ ASes in 120+ countries‣ Web content, mobile content, software updates, etc. ‣ 3 trillion requests on a daily basis

13

IPv4 Address Activity from a CDN

CDN Vantage Point: Active IPv4 Addresses

The CDN Vantage Point

‣ 200,000+ servers in 1500+ ASes in 120+ countries‣ Web content, mobile content, software updates, etc. ‣ 3 trillion requests on a daily basis

‣ 44% active /24 address blocks (6.5M, lower bound raised)‣ 32% active IPv4 addresses (1.2B)

13

IPv4 Address Activity Matrix

130.149.0.6130.149.0.5130.149.0.4130.149.0.3130.149.0.2130.149.0.1

…

…addr

ess

spac

e

days

‣ For each day on which an IP address was active (requested content), we draw a red dot

14

Address Activity Matrix at Scale (“Bacon Strips”)‣ 20K adjacent IP addresses (in active /24s), University Network

time

addr

esse

s

15

Address Activity Matrix at Scale (“Bacon Strips”)

15

‣ Metrics that can capture address activity in space and time‣ Study the effect of addressing mechanisms on‣ Address activity patterns‣ Utilization (seen from the CDN)

time [months]

IP a

ddre

ss a

ctiv

ity w

ithin

/24

0 1 2 3 4

.0.1

27.2

55

time [months]

IP a

ddre

ss a

ctiv

ity w

ithin

/24

0 1 2 3 4

.0.1

27.2

55

time [months]

IP a

ddre

ss a

ctiv

ity w

ithin

/24

0 1 2 3 4

.0.1

27.2

55

University Enterprise ISP Residential ISP

Patterns: Static Address Blocks

Most static address blocks show “activity gaps” 16

time [months]

IP a

ddre

ss a

ctiv

ity w

ithin

/24

0 1 2 3 4

.0.1

27.2

55

time [months]

IP a

ddre

ss a

ctiv

ity w

ithin

/24

0 1 2 3 4

.0.1

27.2

55

time [months]

IP a

ddre

ss a

ctiv

ity w

ithin

/24

0 1 2 3 4

.0.1

27.2

55

DHCP pool US University residential users US ISP residential users DE ISP

Activity/utilization depends on pool size and lease time

Patterns: Dynamic Address Blocks

17

Which Knobs could be adjusted to increase Utilization?

‣ Addressing mechanisms impact address activity

‣ Utilization seen from the CDN:

‣ Static address blocks harbor large supply of potentially unused addresses

‣ Dynamic address blocks could be adjusted to free up underutilized space


Multiplex IPv4:Carrier-Grade NAT Transition to IPv6




ACM IMC ’16 (Best Paper Award)IEEE JSAC ‘16

ContributionMulti-perspective analysis of

address activity, churn, addressing, and utilization.

Findings‣Strong potential for utilization

increase.‣Knobs to adjust: Governance

& Addressing mechanisms.‣Exhaustion effects, stagnation

of routed & active addresses.

Multiplex IPv4 space with Carrier-Grade NAT

130.149.0.1

130.149.0.1

(Carrier-Grade)NAT

18

‣ CGN allows end-user ISPs to ease scarcity issues‣ At the cost of breaking the end-to-end Internet

‣ “Nobody really talks about it”‣ Uncertainty in the community

‣No systematic studies!

Carrier-Grade NAT

19

Research Questions

‣ How can we detect Carrier-Grade NAT?‣ How widespread is Carrier-Grade NAT?‣ What’s the effect on the Internet and its users?

Carrier-Grade NAT

‣ CGN allows end-user ISPs to ease scarcity issues‣ At the cost of breaking the end-to-end Internet

‣ “Nobody really talks about it”‣ Uncertainty in the community

‣No systematic studies!

19

NATs between Subscribers and the Internet

public IPv4

InternetISPSubscriber

public IPv4

internal spacee.g., 192.168.0.0/16 CPE

NAT

NAT44 (subscriber-side)

20


internal spacee.g., 10.0.0.0/8

public IPv4


public IPv4Carrier-GradeNAT


NAT


NAT


NAT44 (carrier-side)

NAT444 (subscriber-side

and carrier-side)

20


internal spacee.g., 10.0.0.0/8

public IPv4


public IPv4Carrier-GradeNAT


NAT


NAT


NAT44 (carrier-side)

NAT444 (subscriber-side

and carrier-side)

20ICSI Netalyzr

BitTorrent DHT

The BitTorrent DHT

Classic BitTorrent Tracker stores peer contact information

tracker

give me peers for torrent XYZ

130.149.1.1:6881130.149.1.2:6882130.149.1.3:6883

BitTorrent DHTPeers store each others’

contact information

give me peers

130.149.1.2:6882130.149.1.3:6883

…

We can use DHT peers as vantage points21

Crawling the BitTorrent DHT

DHTcrawler

give me peers

22

DHTcrawler

i can reach peer 25fc at 130.149.1.2:6881peer 492c at 190.2.0.1:6881…


22

DHTcrawler


NAT

i can reach peer id a82d at 10.53.37.4:6881…

a82d


22

DHTcrawler


NAT

i can reach peer id a82d at 10.53.37.4:6881…

a82d

A B

130.149.1.1:6881 a82d10.53.37.4:6881


A

B

23

BitTorrent Peer Leakage Graph

In this AS: no CGN detected

In this AS: CGN detected

24

Tested with BitTorrent/Netalyzr: 1,791 Eyeball ASes

How widespread is Carrier-Grade NAT Deployment?

25

AFRINICAPNIC

ARINLACNIC

RIPE

0 10 20 30 40 50 60 70

% eyeball ASes covered

AFRINICAPNIC

ARINLACNIC

RIPE

0 5 10 15 20 25

% eyeball ASes CGN−positive

AFRINICAPNIC

ARINLACNIC

RIPE

0 20 40 60 80

100

% cellular ASes CGN−positive

(a) eyeball ASescoverage

(b) eyeball ASesC

GN

-positive(c) cellular ASesC

GN

-positive

‣CGN-positive: 17.1%

‣ particularly in the European and Asia-Pacific Region

Eyeball Networks (Non-Cellular)



25

AFRINICAPNIC

ARINLACNIC

RIPE

0 10 20 30 40 50 60 70


AFRINICAPNIC

ARINLACNIC

RIPE

0 5 10 15 20 25


AFRINICAPNIC

ARINLACNIC

RIPE

0 20 40 60 80

100



(b) eyeball ASesC

GN


GN

-positive

AFRINICAPNIC

ARINLACNIC

RIPE

0 10 20 30 40 50 60 70


AFRINICAPNIC

ARINLACNIC

RIPE

0 5 10 15 20 25


AFRINICAPNIC

ARINLACNIC

RIPE

0 20 40 60 80

100



(b) eyeball ASesC

GN


GN

-positive

‣CGN-positive: 94%

‣ CGN is the norm for cellular




Cellular Networks



25

AFRINICAPNIC

ARINLACNIC

RIPE

0 10 20 30 40 50 60 70


AFRINICAPNIC

ARINLACNIC

RIPE

0 5 10 15 20 25


AFRINICAPNIC

ARINLACNIC

RIPE

0 20 40 60 80

100



(b) eyeball ASesC

GN


GN

-positive

AFRINICAPNIC

ARINLACNIC

RIPE

0 10 20 30 40 50 60 70


AFRINICAPNIC

ARINLACNIC

RIPE

0 5 10 15 20 25


AFRINICAPNIC

ARINLACNIC

RIPE

0 20 40 60 80

100



(b) eyeball ASesC

GN


GN

-positive

‣CGN-positive: 94%

‣ CGN is the norm for cellular




Cellular Networks



CGN is reality for the majority of Internet Users25

What’s the Impact of Carrier-Grade NATs?

private IP1

public IPprivate IP2

private IP3

Internet

NAThome

26

private IP1

public IPprivate IP2

private IP3

Internet

1) Directionality

NAThome


26

Internetinternal IP ranges

external IP ranges

1) Directionality

NATCarrier-Grade


26


external IP ranges

1) Directionality

NATCarrier-Grade


26

IPint, portint IPext, portext


1) Directionality2) Limits/Quotas on flows per subscriber3) Restrictiveness of NAT mappings, timeouts

NATCarrier-Grade


26

external IP ranges

IPint, portint IPext, portext


Deployment Issues‣ Exhaustion of internal IPv4 address space‣ Attribution, host reputation

27



Impact on End Users

‣ Down to 512 ports/subscriber (128 subscribers/IP)

‣ CGN mappings often more restrictive than CPE devices

‣ Restricts (or rules out) peer-to-peer connectivity

27



Impact on End Users

CGN means very different things for different ISPsCGNs limit “how much Internet” subscribers receive

27

‣ Down to 512 ports/subscriber (128 subscribers/IP)

‣ CGN mappings often more restrictive than CPE devices

‣ Restricts (or rules out) peer-to-peer connectivity


Multiplex IPv4:Carrier-Grade NAT

ACM IMC ’16 (IRTF ANRP Award)

Transition to IPv6

ContributionFirst broad and systematic study

of CGN deployment in the Internet and properties.



Findings‣CGNs are very broadly

deployed (majority of users).

‣CGNs directly limit end-users’ connectivity and resources.

‣CGN deployment issues (internal space, attribution,..).









Strategy (iii): Transition to IPv6

‣ IPv6 (est. 1998) comes with 128-bit IP addresses

‣ Long-term solution to the IPv4 scarcity problem

‣ Enormous Task: Replacing the central Internet Protocol

28

(iii) CPE

(i) OS(ii) applications

Home network Dual-stack ISP Service providers

IPv4 traffic

IPv6 traffic

(iv) ISP connectivity (v) service availability

Inte

rnet

Strategy (iii): Transition to IPv6

‣ IPv6 (est. 1998) comes with 128-bit IP addresses

‣ Long-term solution to the IPv4 scarcity problem

‣ Enormous Task: Replacing the central Internet Protocol

28

(iii) CPE

(i) OS(ii) applications

Home network Dual-stack ISP Service providers

IPv4 traffic

IPv6 traffic

(iv) ISP connectivity (v) service availability

Inte

rnet

‣ As of 2017:‣ A minority of Internet hosts speak IPv6‣ Majority of Internet traffic carried over IPv4












Transition to IPv6

ACM IMC ’14, PAM ’15, PAM ‘17

ContributionAnalysis of IPv4/IPv6

connectivity, traffic components, and interplay.

Findings‣ IPv6 connectivity increases,

yet lags behind IPv4.

‣Traffic over IPv6 lags behind connectivity.

‣Barriers for IPv6 traffic (devices, software, networks).









IPv4 Exhaustion: An Unprecedented Problem

‣ IPv4 addresses are truly global virtual resources

‣ No central authority, independent decisions

29




29

‣ Looming IPv4 exhaustion was recognized early (~1990)

‣ Yet, IPv4 supplies lasted until ~2011




29

‣ Looming IPv4 exhaustion was recognized early (~1990)

‣ Yet, IPv4 supplies lasted until ~2011

‣ Today: Economic pressure due to IPv4 scarcity!

‣ Growing IPv4 address markets

‣ Widespread Carrier-Grade NAT deployment

‣ Increasing Dual-Stack IPv4/IPv6 deployment






Transition to IPv6

ACM IMC ’14, PAM ’15, PAM ‘17





ContributionAnalysis of IPv4/IPv6

connectivity, traffic components, and interplay.

Findings‣ IPv6 connectivity increases,

yet lags behind IPv4.

‣Traffic over IPv6 lags behind connectivity.

‣Barriers for IPv6 traffic (devices, software, networks).











empirical analysis of the effects and the mitigation of ...empirical analysis of the effects and the...

Documents