dataset comparison - dns-oarc (indico) · bradley huffaker, marina fomenkov, and kc claffy oarc...
TRANSCRIPT
Bradley Huffaker, Marina Fomenkov, and kc claffy
OARC Fall 2014 Workshop October 12, 2014
Dataset Comparison IPv4 vs IPv6 traffic seen at DNS root servers
IPv4 exhaustion
2
RIR
Add
ress
Poo
l (/8
s)
http://www.potaroo.net/tools/ipv4/
For most practical purposes IPv4 exhaustion is here.
goal
3
We compare country level allocation and observed deployment of IPv4 and IPv6 by examining DNS root traffic collected by DNS-OARC members as part of the Day In The Life (DITL) measurements.
http://www.dns-oarc.net/oarc/data/ditl
datasets• DNS root traces!
DNS-OARC Day In The Life (DITL)!!
• IP allocated blocks!RIR delegation files!also used for country-level geolocation!!
• IP announced prefixes!BGP tables collected by Route Views Project and RIPE NCC RIS!!
• Gross Domestic Product (GDP)!World Bank
4
http://www.caida.org/research/policy/dns-country/Number of Countries
Almost all countries now
have at least one IPv6
allocation and a growing
fraction were seen at the
DNS roots.
5
IPv4 IPv6
2009
2010
2011
2012
2013
2014
number of countries at least one allocation
http://www.caida.org/research/policy/dns-country/interactive web site
6
Allows users to switch between set views.
presets
manual controlAllows full control over both axis and node size
manual control switch
http://www.caida.org/research/policy/dns-country/fraction of blocks observed
Fraction of address blocks seen at the roots much tighter for IPv4 and IPv6 address blocks. Reflects more uniform usage
of IPv4 addresses. 7
IPv6
IPv4
countries with zero
China
U.S.A
Germany
India
Brazil
1e+0 2e+0 3e+0 1e+1 2e+1 3e+1 1e+2 2e+2 3e+2 1e+3 2e+3 3e+3
2014 IPv6 blocks allocated by RIR (blocks)1e+0
2e+03e+04e+0
1e+1
2e+13e+14e+1
1e+2
2e+23e+24e+2
2014
IPv6
blo
cks s
een
at th
e D
NS
root
s (bl
ocks
)
Russia
Saint Kitts and Nevis
U.S.A
Canada
1e+0 2e+0 1e+1 2e+1 1e+2 2e+2 1e+3 2e+3 1e+4 2e+4
2014 IPv4 blocks allocated by RIR (blocks)1e+02e+0
1e+12e+1
1e+22e+2
1e+32e+3
1e+4
2014
IPv4
blo
cks s
een
at th
e D
NS
root
s (bl
ocks
)
http://www.caida.org/research/policy/dns-country/
Russia
U.S.A
JapanFrance
Canada
Brazil
Monaco1e+0 2e+0 1e+1 2e+1 1e+2 2e+2 1e+3 2e+3 1e+4 2e+4
2014 IPv4 blocks allocated by RIR (blocks)1e+0
2e+03e+0
1e+1
2e+13e+1
1e+2
2e+23e+2
1e+3
2e+33e+3
2014
IPv6
blo
cks a
lloca
ted
by
RIR
(blo
cks)
Nigeria
Australia
Greece
Ukraine
Netherlands
China
U.S.A
Japan
Egypt
France
Germany
BrazilMonaco
Great Britain
1e+0 2e+0 1e+1 2e+1 1e+2 2e+2 1e+3 2e+3 1e+4 2e+4
2009 IPv4 blocks allocated by RIR (blocks)1e+0
2e+03e+0
1e+1
2e+13e+1
1e+2
2e+23e+2
1e+3
2e+33e+3
2009
IPv6
blo
cks a
lloca
ted
by
RIR
(blo
cks)
ratio IPv4 to IPv6 blocks
Only Monaco went to zero. Brazil went from unusually low ratio of IPv4 to IPv6, to higher then normal. Overall the ratio of
IPv4 and IPv6 allocations was more similar in 2014.8
2009
2014MonacoBrazil
http://www.caida.org/research/policy/dns-country/
China
Hong Kong
India
U.S.A
VietnamCroatia
France
Germany
1e+0 2e+0 1e+1 2e+1 1e+2 2e+2 1e+3 2e+3 1e+42014 IPv6 addresses seen by the DNS roots (addresses)
1e+0
1e+1
1e+2
1e+3
1e+4
1e+5
1e+6
2014
IPv6
pac
kets
seen
by
the
DN
S ro
ots (
pack
ets)
packets and addresses
Similar with other metrics, IPv4 is less scattered than IPv6.9
IPv6
IPv4
radius: packetsaddresses
Saint Vincent & GrenadinesHaiti
AlbaniaNigeria
Philippines
South Korea
ChinaU.S.A
Germany
1e+0 2e+0 1e+1 2e+1 1e+2 2e+2 1e+3 2e+3 1e+4 2e+4 1e+5 2e+5
2014 IPv4 addresses seen by the DNS roots (addresses)
1e+1
1e+2
1e+3
1e+4
1e+5
1e+6
1e+7
2014
IPv4
pac
kets
seen
by
the
DN
S ro
ots (
pack
ets)
http://www.caida.org/research/policy/dns-country/ratio packets / addresses
The countries with the largest ratios tend to have fewer addresses. China has 40% of US IPv6 addresses, but 80% of
it’s IPv6 packets.10
IPv4
radius: packetsaddresses
China
Hong Kong
India
U.S.A
Vietnam
France
Germany
1e+0 2e+0 1e+1 2e+1 1e+2 2e+2 1e+3 2e+3 1e+4
2014 IPv6 addresses seen by the DNS roots (addresses)1e+0
2e+03e+0
1e+1
2e+13e+1
1e+2
2e+23e+2
1e+3
2014
IPv6
num
ber o
f pac
kets
seen
by
the
DN
S ro
ots p
er a
ddre
ss (p
acke
ts/ad
dres
s)
Saint Vincent & Grenadines
HaitiAlbania
Nigeria
South KoreaSaint Kitts and Nevis China
Hong KongIndia
U.S.A
Germany
1e+0 2e+0 1e+1 2e+1 1e+2 2e+2 1e+3 2e+3 1e+4 2e+4 1e+5 2e+52014 IPv4 addresses seen by the DNS roots (addresses)
2e+0
1e+1
2e+1
1e+2
2e+2
1e+3
2e+3
1e+4
2014
IPv4
num
ber o
f pac
kets
seen
by
the
DN
S ro
ots p
er a
ddre
ss (p
acke
ts/a
ddre
ss)
IPv6
http://www.caida.org/research/policy/dns-country/Gross Domestic Product
There is a positive correlation between wealthier countries and the ratio of IPv4 blocks seen at the DNS roots, but not
IPv6 blocks.11
IPv6
IPv4
radius: IPv6 allocated blocks
radius: IPv4 allocated blocks
U.S.A
Canada
Russia
Australia
Germany
Brazil
China
Japan
Saint Kitts and Nevis
Solomon Islands
TurkmenistanEritrea
1e+8 1e+9 1e+10 1e+11 1e+12 1e+132014 GDP [Gross Domestic Product] (US$)
9e+01e+1
2e+1
3e+1
4e+1
5e+16e+17e+18e+19e+11e+2
2014
Rat
io o
f IPv
4 bl
ocks
seen
at t
he D
NS
root
s to
allo
cate
d (%
)
U.S.A
Brazil China
Japan
LiberiaMicronesia
1e+8 1e+9 1e+10 1e+11 1e+12 1e+13
2014 GDP [Gross Domestic Product] (US$)
2e+0
3e+04e+05e+0
1e+1
2e+1
3e+14e+15e+1
1e+2
2014
Rat
io o
f IPv
6 bl
ocks
seen
at t
he D
NS
root
s to
allo
cate
d (%
)
conclusion• number of countries with allocations!
IPv4 allocations almost 100% seen!IPv6 allocations have gone from 50% to 61% seen!Almost all countries now have at least one for both!!
• IPv6 fraction of blocks seen more dissimilar than IPv4!!
• correlation between number of IPv4 and IPv6 blocks increased between 2009 and 2014!!
• negative correlation between wealth and ratio of blocks seen in IPv4, but not IPv6
12
http://www.caida.org/research/policy/dns-country/