Transcript
Letter From the Editor
After the success of 2011’s World IPv6 Day event, the Internet Society organized the World IPv6 Launch event in June
2012, promoting it with the tag line “This Time It Is For Real”. The associated Web site (http://www.worldipv6launch.
com) noted “Major Internet service providers (ISPs), home networking equipment manufacturers, and Web companies
around the world are uniting to redefine the global Internet and permanently enable IPv6 for their products and
services on 6 June 2012.” By all indications, this year’s event was a success as well, and it came none too soon, as
September saw RIPE, the European Internet registry, announce that they were down to their final “/8” block (~16
million) of IPv4 addresses, and ARIN, the regional Internet registry for the Americas, announce that they only had
three “/8” blocks remaining. It is critical that service providers, equipment manufacturers, and content providers
continue planning for a long-term transition to IPv6 while enabling support for it now.
The state of broadband connectivity in the United States also continued to be a key area of government focus during
the second quarter. The United States Federal Communications Commission (FCC) released a report that examined
how the country’s ISPs fared when actual download speeds were compare to advertised download speeds. In addition,
President Obama signed an executive order intended to “ensure that agencies charged with managing Federal properties
and roads take specific steps to adopt a uniform approach for allowing broadband carriers to build networks
on and through those assets and speed the delivery of connectivity to communities, businesses, and schools.” The
White House also announced the “US Ignite” partnership between cities, corporate and non-profit entities, and
national research universities, charged with developing new services to take advantage of high-speed broadband
networks. At a state level, based on interactions that I have had with various folks working on “local” efforts, a lot
of attention is being paid to connection speeds and broadband adoption within the states (and how to improve
it, of course), not to mention concern about how a given state compares to other similar/neighboring states.
Going forward, as IPv6 adoption increases, broadband connectivity (in the U.S. and around the world) continues to
improve, and new applications are developed to take advantage of this improved connectivity, Akamai stands ready
to help customers, partners, and end users exploit these advantages to their fullest potential, ensuring optimal delivery
speeds and unmatched scalability. And with this increased usage will come a wealth of additional data that we
will be able to aggregate, analyze, and present in outlets such as the State of the Internet report and Akamai IO.
As always, if you have questions, comments, or suggestions about the State of the Internet report, connect with
us via e-mail at stateoftheinternet@akamai.com, or on Twitter at @akamai_soti.
.David Belson
Table of Contents Table of Contents
ExEcutivE Summary
3
SEction 1: SEcurity
4
1.1 Attack Traffic, Top Originating Countries
4
1.2 Attack Traffic, Top Ports
4
1.3 SSL Insight, Client-Side Ciphers
1.4 Password Hash Disclosures
6
SEction 2: intErnEt PEnEtration 7
2.1 Unique IPv4 Addresses 7
2.2 IPv4 Exhaustion 8
2.3 IPv6 Adoption 9
2.4 World IPv6 Launch
2.5 New Generic Top Level Domains (gTLDs) 11
SEction 3: GEoGraPhy.Global 12
3.1 Global Average Connection Speeds 12
3.2 Global Average Peak Connection Speeds 13
3.3 Global High Broadband Connectivity 14
3.4 Global Broadband Connectivity
SEction 4: GEoGraPhy .unitED StatES 16
4.1 United States Average Connection Speeds 16
4.2 United States Average Peak Connection Speeds 17
4.3 United States High Broadband Connectivity 18
4.4 United States Broadband Connectivity 19
SEction 5: GEoGraPhy.aSia Pacific rEGion
5.1 Asia Pacific Average Connection Speeds 20
5.2 Asia Pacific Average Peak Connection Speeds 21
5.3 Asia Pacific High Broadband Connectivity 22
5.4 Asia Pacific Broadband Connectivity 22
SEction 6: GEoGraPhy.EmEa 24
6.1 EMEA Connection Speeds 24
6.2 EMEA Average Peak Connection Speeds
6.3 EMEA High Broadband Connectivity 26
6.4 EMEA Broadband Connectivity
27
SEction 7: mobilE connEctivity
28
7.1 Connection Speeds on Mobile Networks
28
7.2 Mobile Browser Usage Data
7.3 Mobile Traffic Growth As Observed By Ericsson
31
7.4 Traffic Variation by Screen Resolution as Observed by Ericsson
32
SEction 8: intErnEt DiSruPtionS
33
8.1 China 33
8.2 Syria 34
SEction 9: aPPEnDix
SEction 10: EnDnotES
36
ⓒ 2012 Akamai Technologies, Inc. All Rights Reserved
Executive SummaryExecutive Summary Executive Summary
Akamai’s globally distributed Intelligent Platform allows us to gather massive amounts
of information on many metrics, including connection speeds, attack traffic, network
connectivity/availability/latency problems, and IPv6 growth/transition progress, as well
as traffic patterns across leading Web sites and digital media providers. Each quarter,
Akamai publishes the State of the Internet report. This report includes data gathered
from across the Akamai Intelligent Platform during the second quarter of 2012 about
attack traffic, broadband adoption, and mobile connectivity, as well as trends seen
in this data over time. In addition, this quarter’s report includes insight into SSL, the
state of IPv6 adoption as measured by Hurricane Electric and the World IPv6 Launch
event, and observations from Akamai partner Ericsson regarding variations observed
in mobile traffic patterns by screen resolution and screen size.
Security
During the second quarter of 2012, Akamai observed attack
traffic originating from 188 unique countries/regions. China
remained the top attack traffic source, once again responsible
for 16% of total observed attack traffic. The United States
and Turkey held the second and third place spots respectively,
accounting for just under 20% of observed attack traffic
combined. Attack traffic concentration declined from the first
quarter of 2012, with the top 10 ports seeing 62% of observed
attack traffic. In June, unnamed attackers disclosed nearly 10.5
million passwords after compromising three leading Web sites
that were not using a technique known as “salted hashing”
for securely storing encrypted passwords, meaning that the
compromised password files were much more vulnerable than
they should have been.
Internet and Broadband Adoption
Akamai observed a 0.1% quarterly decrease in the number
of unique IPv4 addresses connecting to Akamai, falling to just
over 665 million, just less than one million fewer addresses than
were seen in the first quarter. Looking at connection speeds, the
global average connection speed grew 13% to 3.0 Mbps, and
the global average peak connection speed grew 19% to 16.1
Mbps. At a country level, South Korea had the highest average
connection speed at 14.2 Mbps, while Hong Kong recorded
the highest average peak connection speed, at 49.2 Mbps.
As was noted in last quarter’s report, Akamai is now defining
“high broadband” as connections of 10 Mbps or higher and
“broadband” as connections of 4 Mbps or higher. Globally, high
broadband (>10 Mbps) adoption dropped 1.6% in the second
quarter, staying at 10%, and South Korea continued to have
the highest level of high broadband adoption, at 49%. Global
broadband (>4 Mbps) adoption dropped 2.8% to 39%, with
South Korea having the highest level of broadband adoption,
at 84%. Note that starting with last quarter’s report, we are
no longer including figures for narrowband (<256 kbps) adoption,
nor city-level data.
Mobile Connectivity
In the second quarter of 2012, average connection speeds on
known mobile network providers ranged from a high of 7.5
Mbps down to 340 kbps. Average peak connection speeds for
the quarter ranged from 44.4 Mbps down to 2.5 Mbps. Based
on data collected by Ericsson, mobile data traffic doubled from
the second quarter of 2011 to the second quarter of 2012,
and grew 14% quarter-over-quarter.
Analysis of Akamai IO data collected in June of a sample of
requests to the Akamai Intelligent Platform indicates that for
users of mobile devices on cellular networks, the largest percentage
of requests (~38%) comes from Android Webkit, with Apple’s
Mobile Safari close behind (~33%). However, for users of
mobile devices across all networks, Apple’s Mobile Safari
accounts for approximately 60% of requests, indicating that
significantly more users of iOS devices use these devices
on Wi-Fi networks .heavily driven by iPad usage.
3
SECTION 1:
Security
SECTION 1:
Security
Akamai maintains a distributed set of agents deployed across the Internet that monitor
attack traffic. Based on data collected by these agents, Akamai is able to identify the top
countries from which attack traffic originates, as well as the top ports targeted by these
attacks. (Ports are network-level protocol identifiers.) This section provides insight into
port-level attack traffic, as observed and measured by Akamai, during the second quarter
of 2012. It also provides insight into trends related to the usage of client-side ciphers
for SSL connections to Akamai, as well observations on password hash disclosures that
occurred on a number of large Web sites during the second quarter.
1.1 Attack Traffic, Top Originating Countries
During the second quarter of 2012, Akamai observed attack
traffic originating from 188 unique countries/regions, up from
182 in the prior quarter. As shown in Figure 1, China remained
the source of the largest volume of observed attack traffic,
accounting for approximately 16% of the total, consistent
with the first quarter. The United States saw a slight quarterly
increase, originating 12% of observed attacks in the second
quarter. Nine of the top 10 countries remained consistent
quarter-over-quarter, with the exception of Germany, which
ceded its place on the list to Italy this quarter. Six of the top
10 countries saw quarterly growth in the associated percentage
of observed attack traffic, while three saw a quarterly decline.
In examining the regional distribution of observed attack traffic
in the second quarter, we found that nearly 38% originated in
the Asia Pacific/Oceania region, just over 36% in Europe, 23%
in North and South America, and just under 3% from Africa.
The Asia Pacific/Oceania region was the only one where attack
traffic concentration declined quarter-over-quarter.
1.2 Attack Traffic, Top Ports
As shown in Figure 2, attack traffic concentration among the
top 10 ports declined during the second quarter of 2012, with
these ports responsible for 62% of observed attacks, down
from 77% last quarter, and consistent with the level seen in the
fourth quarter of 2011. It appears that this decline is largely
attributable to the significant decline in the percentage of attacks
targeting Port 445, after an unusually large increase last quarter.
In addition to the decrease seen in the percentage of attacks
targeting Port 445, decreases were also seen for Port 23, Port
1433, Port 3389, Port 80, Port 22, and Port 4899. The average
relative decline seen across these ports was on the order of
25%. Port 8080 saw the greatest increase quarter-over-quarter,
jumping over 200% (but still the target of less than 2%
of observed attacks), with Port 135 and Port 139 also seeing
quarterly increases. Research does not indicate the discovery of
any new attacks or vulnerabilities during the quarter that would
account for the doubling of Port 8080-targeted attacks.
4
1
8
figure 1: Attack Traffic, Top Originating Countries
1 China 16% 16%
2 United States 12% 11%
3 Turkey 7.6% 5.7%
4 Russia 6.3% 7.0%
5 Taiwan 5.4% 5.3%
6 Brazil 4.6% 4.0%
7 Romania 3.5% 3.0%
8 India 2.9% 3.0%
9 Italy 2.1% 1.9%
10 South Korea 2.1% 4.3%
. Other 37% 39%
Q1 ‘12 % Q2 ‘12 % traffic country
2
7
6
3
9
5
10
4
figure 2: Attack Traffic, Top Ports
445 Microsoft-DS 32% 42%
23 Telnet 9.2% 11%
1433 Microsoft SQL Server 4.5% 4.9%
3389 Microsoft Terminal Services 4.2% 4.6%
80 WWW (HTTP) 3.8% 5.0%
22 SSH 2.2% 3.4%
8080 HTTP Alternate 1.9% 0.6%
135 Microsoft-RPC 1.9% 1.6%
4899 Remote Administrator 1.2% 1.6%
139 NetBIOS 1.0% 0.8%
Various Other 38% .
Q1 ‘12 % Q2 ‘12 % traffic Port use Port
Port 445 remained the most targeted port in eight of the top 10
countries, accounting for as many as 85 times (in Romania) the
number of attacks seen by the next most targeted port. Once
again, Port 23 remained the most targeted port in observed
attacks originating in Turkey, with seven times as many attacks
targeting that port than Port 445, the next most targeted port.
In China, Port 1433 remained the most targeted port, with 1.7
times as many attacks targeting that port as Port 3389, the
next most targeted port for attacks observed to be originating
from the country. Port 23 was the most common second-most
targeted port, ranking second in India, South Korea, Taiwan,
and the United States potentially indicating the prevalence of
malware in these countries that attempts to exploit default or
common passwords on remotely accessible systems that would
allow attackers to gain access to these systems.
1.3 SSL Insight, Client-Side Ciphers
In addition to the large number of requests for content that
Akamai serves over HTTP (Port 80), the Akamai Intelligent
Platform also services millions of requests per second for secure
content over HTTPS/SSL (Port 443). This massive volume of
encrypted traffic provides Akamai with a unique perspective
on the client-side SSL ciphers that are in popular use, as well
as their usage trends over time. The statistics presented in this
section are for SSLv3 and TLSv1.
Figure 3 illustrates the distribution of SSL ciphers presented by
Web clients (generally browsers) to Akamai’s Secure Content
Delivery Network during the second quarter of 2012. Once
again, the shifts in usage trends varied from those observed in
prior quarters. As shown in the figure, it appears that usage of
the RC4-MD5-128 cipher grew significantly during the quarter,
Microsoft SQL Sever 4.5%
Microsoft Terminal Services 4.2%
WWW (HTTP) 3.8%
SSH 2.2%
HTTP Alternate 1.9%
Microsoft RPC 1.9%
Remote Administrator 1.2%
NetBIOS 1.0%
Telnet
9.2%
Other 38%
Microsoft-DS 32%
including an unusual bump seen throughout May. Usage of
this cipher increased from 10.3% at the start of the quarter, to
14.8% at the end of the quarter .an increase of 44%. Usage
of other ciphers declined across the course of the quarter, with
RC4-SHA-128 losing the most, declining from 3.7% to 3.2%.a loss
of just over 14%. Usage of AES-256-SHA-1 once again declined
slightly, losing 2.9% to end the quarter at 43.8% usage.
AES128-SHA-1 also lost some ground in the second quarter,
dropping 6.6% to 36.3% usage. Despite the declines, these
two ciphers are still responsible for 80% of the ciphers presented
to Akamai servers.
100%
90%
80%
70%
60%
50%
40%
30%
20%
10%
0%
AES256-SHA-1
RC4-MD5-128
DES-CBC3-SHA-168
AES128-SHA-1
RC4-SHA-128
figure 3: Client-Side SSL Ciphers Observed by Akamai, Q2 2012
1-Jan
8-Jan
15-Jan
22-Jan
29-Jan
5-Feb12-Feb19-Feb26-Feb4-Mar11-Mar18-Mar25-Mar
5
SECTION 1:
Security (continued)
SECTION 1:
Security (continued)
On the server side, a nonprofit organization known as the Trust
worthy Internet Movement has started publishing the results of
scans of the HTTPS implementations on Web sites included in
Alexa’s list of the top one million sites, in an initiative known as
“SSL Pulse”.1 The idea behind the SSL Pulse initiative, according
to its Web site, is to focus on auditing the SSL ecosystem, raising
awareness, and providing tools and documentation to Web
site owners so they can improve their SSL implementations. We
hope to include data and observations from SSL Pulse’s measure
ments in upcoming issues of the State of the Internet report
in order to provide a more comprehensive picture of the state
of SSL usage on both Web clients and Web servers.
1.4 Password Hash Disclosures
On June 6th, 2012, it was discovered that 6.5 million password
hashes from social networking site LinkedIn had been revealed
by hackers, and that some 300,000 of these hashes had already
been compromised. Later the same day, it was revealed that 1.5
million password hashes from online dating site eHarmony had
also been posted to the Internet. The final member of this list,
music site Last.fm, revealed that it had discovered a file containing
2.5 million password hashes of their own the week before. In
total, unnamed attackers had disclosed nearly 10.5 million passwords
from these three companies in the span of a week.
A hash, or message digest, is a one-way encryption algorithm
that allows the original data to be verified, but makes it impossible
to decrypt to find the original value of the data. In other words,
a password can be hashed, and when the password is used
again, it can be verified against the hash, but it is impossible
to decrypt the hash to find the original password.
Hashes are used extensively on Web sites to protect passwords
and maintain the integrity of the password files without exposing
the actual password. Hashes can be broken, but to do so,
requires methods such as dictionary or brute force attacks that
hash words and random characters in order to find collisions
with the hashed data. Dictionary attacks use common words
in the hash function to find the collisions with the encrypted
passwords and are quite fast. Brute force attacks use strings of
random characters to find collisions with stronger passwords,
but can be quite time and computing power intensive.
Examples of commonly used hashing algorithms include ‘Secure
Hashing Algorithm 1’ or SHA-1, used by LinkedIn, and the
slightly older and less secure Message Digest 5 or MD5, used by
Last.fm and eHarmony. A well-implemented password hashing
function includes what is called a ‘salt’, which is a series of
random characters that are prepended to the password in order
to greatly increase the amount of time and computing power
required to find collisions with the hashed passwords. Unfortunately,
none of the companies whose passwords were compromised
were using salted hashes, meaning their compromised password
files were much easier to find collisions in than they should
have been.
All three companies have since implemented salt in their hashing
functions for passwords, and have implemented ‘additional
security features’, though the exact nature of these security
measures has yet to be disclosed. One of the largest concerns
with these compromises is that many users re-use passwords
across many sites, for example using the same password for a
site like eHarmony as they use for their banking site. The
exposure of the password for one site may lead to a compromise
of a completely unrelated site due to bad end-user
password practices. Users of the affected sites were notified
and asked to change their passwords, but many who are
already overloaded by e-mail may have missed the notifications,
or assumed the notifications were simply spam.
An ancillary concern from these compromises is the amount
of user information that may have been compromised along
with the passwords. Both LinkedIn and eHarmony have large
amounts of very personal data about their users, and this data
can be used to craft highly targeted phishing campaigns
or to answer user security questions on other sites.
None of the three companies, LinkedIn, Last.fm or eHarmony,
have disclosed much information on the nature of how they
were compromised. Security researchers and hackers have
used various tools to discover all of the passwords that were
contained in these files, so it is important for users to verify that
they have changed their passwords. Since password re-use is
common, it is important for users to look at using password
vault software to help create and store strong passwords, rather
than reusing the same ones repeatedly.
6
SECTION 2:
Internet Penetration
SECTION 2:
Internet Penetration
2.1 Unique IPv4 Addresses
Through its globally-deployed Intelligent Platform, and by virtue
of the approximately two trillion requests for Web content that
it services on a daily basis, Akamai has unique visibility into
levels of Internet penetration around the world. In the second
quarter of 2012, over 665 million IPv4 addresses, from 242
countries/regions, connected to the Akamai Intelligent
Platform.0.1% fewer than in the first quarter of 2012 and
10% more than in the second quarter of 2011. Although we
see more than 600 million unique IPv4 addresses, Akamai
believes that we see well over one billion Web users. This is
because, in some cases, multiple individuals may be represented
by a single IPv4 address (or a small number of IPv4 addresses),
because they access the Web through a firewall or proxy server.
Conversely, individual users can have multiple IPv4 addresses
associated with them due to their use of multiple connected
devices. Unless otherwise specified, the use of “IP address”
within Section 2.1 refers to IPv4 addresses.
As shown in Figure 4, the global unique IP address count saw
an unusual, though extremely slight, decline as compared to
the prior quarter . just less than one million fewer addresses.
This quarterly decline is in contrast to the quarterly increase in
unique IP address count that we have come to expect over the
last four years of this report. However, we do not believe that
this decline portends any sort of imminent decline in Internet
penetration globally, nor do we believe that it is yet related to
increased content consumption over IPv6. Rather, we believe
that the decline is likely due to changes made to Akamai’s back-end
data collection and analysis processes. Along these lines,
nominal quarterly declines were also seen among four of the
top 10 countries in the second quarter, with the United States,
Japan, South Korea, and Russia all having lower unique IP address
counts. Increases were seen in the remaining six countries, with
Brazil again seeing very strong growth, adding over 2 million IP
addresses quarter over quarter. Globally, quarterly growth was
seen in just over 60% of countries/regions around the world,
with 31 countries seeing increases of 10% or more.
Looking at year-over-year changes, we see that among the
top 10 countries, seven countries had higher unique IP address
counts as compared to the second quarter of 2011, with five
of the seven increasing more than 10%. In contrast, the United
States saw a slight yearly decline, while Japan and South Korea
both continued the string of losses observed over the last several
quarters. As noted in previous quarters’ reports, the longer-term
negative trending in Japan and South Korea could be due to a
number of possible causes, including ongoing changes to data
in Akamai’s EdgeScape IP geolocation database, shifts in IP
address block utilization by local network service providers,
increased use of proxies, or deployment of so-called “large
scale NAT” (network address translation) infrastructure by
carriers in an effort to conserve limited IPv4 address space. On
a global basis, however, approximately 80% of countries/
regions had higher unique IP address counts year-over-year.
Among those countries/regions that saw declines, the largest
losses were generally seen in geographies with comparatively
smaller address counts.low enough that they do not qualify
for inclusion in subsequent sections.
. Global 665,180,961 -0.1% 10%
1 United States 142,879,594 -2.4% -0.3%
2 China 93,604,214 1.3% 22%
3 Japan 39,876,417 -1.6% -11%
4 Germany 36,196,309 0.4% 3.7%
5 United Kingdom 26,579,255 3.3% 17%
6 France 26,103,462 2.2% 7.4%
7 Brazil 21,546,894 12% 40%
8 South Korea 19,666,997 -0.7% -14%
9 Italy 17,965,986 6.2% 25%
10 Russia 15,472,955 -2.4% 26%
yoy
change
QoQ
change
Q2 ‘12 unique
iP addresses
country
4
7
2
10
3
8
6
5
1
9
figure 4: Unique IPv4 Addresses Seen By Akamai
7
SECTION 2:
Internet Penetration (continued)
SECTION 2:
Internet Penetration (continued)
2.2 IPv4 Exhaustion
The number of available IPv4 addresses continued to decline
during the second quarter of 2012, as Regional Internet Registries
(RIRs) continued to assign/allocate blocks of address space to
requesting organizations within their respective territories.2
Based on data published by the RIRs,3 Figure 5 compares IPv4
address assignment/allocation activity by RIR during the second
quarter of 2012.
As shown, both RIPE and APNIC had fairly steady consumption
patterns during the second quarter. However, ARIN, LACNIC,
and AFRINIC all saw “stairsteps” occur in their consumption,
where comparatively larger numbers of IPv4 addresses were
assigned/allocated on one or more days during the quarter. The
most pronounced step in the graph occurred with ARIN on
May 15, when over 13 million IPv4 addresses were assigned/
allocated in three separate transactions. The largest involved
over 11 million addresses, assigned in multiple smaller blocks to
Bell Northern Research, according to ARIN records.4 Vodafone
also got a block of just over one million addresses that day as
well.5 The second step evident on the ARIN graph is from
an allocation of just over two million IPv4 addresses to T-Mobile
USA.6 At LACNIC, the evident stairstep on April 25 appears
to be due to the allocation of just over two million addresses
to Vivo S.A.7 An allocation of just over one million addresses
to Vodacom on June 6 accounts for the stairstep observed in
AFRINIC’s otherwise minimal consumption of IPv4 space in the
second quarter.8
25
Millions of IPv4 Addresses
20
15
10
5
0
In late May, Daniel Karrenberg, RIPE’s Chief Scientist noted9
in a post to industry forum CircleID that the RIPE NCC’s pool
of unallocated IPv4 addresses is expected to reach the “last
/8” this year, which means that the organization will have
16,777,216 IPv4 addresses left in the available pool. When that
milestone is reached, RIPE will go into an “austerity mode”
similar to the one that APNIC has been operating under since
the second quarter of 2011, in which RIPE will only be able
to distribute IPv6 addresses and a one-off allocation of IPv4
address space (1,024 IPv4 addresses) from the “last /8” to
those members that meet the policy requirements. One may
expect that the imminent arrival of such a milestone would
cause a rush to obtain IPv4 address space before small one-off
allocations become the norm. However, in examining the yearly
volume of IPv4 address distribution in the RIPE service region,
it was found that consumption during the first four months
of 2012 was in line with distribution trends for the prior three
years, and that there was no evidence of a “last-minute rush”
taking place during the early part of the year. Regarding this
observation, Karrenberg noted, “It’s good to see that the
address space distribution policies set by the RIPE community to
ensure that the remaining IPv4 address space is conserved and
distributed fairly over the last few years have worked well.”
(It is worth noting that according to the RIPE Web site, “On Friday
14 September, 2012, the RIPE NCC, the Regional Internet
Registry (RIR) for Europe, the Middle East and parts of Central
Asia, distributed the last blocks of IPv4 address space from the
available pool.”10 This means that RIPE has, indeed, reached
the last “/8” of their address space.)
AFRINIC
APNIC
ARIN
LACNIC
RIPENCC
1-Apr8-Apr15-Apr22-Apr29-Apr5-May13-May20-May27-May3-June10-June17-June24-June
figure 5: Total Number of IPv4 Addresses Assigned/Allocated by RIRs, Q2 2012
8
2.3 IPv6 Adoption
As Akamai continues to roll out IPv6 support across its solution
portfolio, we will endeavor to include data in the State of the
Internet report on IPv6 adoption based on the analysis of IPv6
requests to, and traffic delivered by, the Akamai Intelligent
Platform. However, until such time as we can include comprehensive
Akamai data on IPv6 adoption, we will continue to look
to third-party data.
One helpful source of IPv6 information is Hurricane Electric,
which provides Internet transit, Web hosting, and collocation
services and notes that it is “considered the largest IPv6 backbone
in the world as measured by number of networks connected.”11
A white paper12 available from Hurricane Electric notes that
it has operated IPv6 network elements since 2000 and that it
implemented extensive native IPv6 peering in early 2006 as a
result of a core router and backbone upgrade. Hurricane Electric
also publishes the output of a set of measurement tools on its
“Global IPv6 Deployment Progress Report” page, available at
http://bgp.he.net/ipv6-progress-report.cgi.
Figure 6 illustrates the growth in the number of ASes in the
global IPv6 routing table during the second quarter of 2012,
comparing it to the second quarters of the previous three years
as well. As shown in the figure, the second quarter 2012 rate
of growth was just under 7%, lower than the growth rates
seen in 2009, 2010, or 2011. Just under half as many ASes
were added in the second quarter of 2012 as in 2011, though
it was over three times as many as was added in 2009, and
almost twice as many as in 2010. It is likely that the significant
growth seen in 2011 was related to preparations for World
IPv6 Day.many ASes were likely already IPv6 capable ahead
of 2012’s World IPv6 Launch event. In addition, as has been
discussed in the past, while the “IPv6 ASes” metric provides
some perspective around IPv6 adoption, it is also important to
recognize that not all autonomous systems are equivalent. That
is, IPv6 adoption on an autonomous system that is associated
with a large number of end users/subscribers is ultimately more
meaningful and impactful for measuring the ultimate success
of IPv6 than adoption by an autonomous system that is not
directly associated with end user connectivity/traffic.
To that end, in a May presentation13 at the “ENOG 3” (Eurasia
Network Operators’ Group) conference, Jim Cowie of Internet
monitoring firm Renesys noted that “less than 15% of ASNs
worldwide participate”, with only 5.78% of those in the Eurasian
region participating. He noted that there has been success in
“convincing a small number of large ASNs”, but that there was
still work to be done in convincing “large numbers of small
ASNs to implement IPv6”. As one large ASN example, in April,
Comcast announced that it had made IPv6 service available to
residential users in two U.S. cities that were using one of six
specific IPv6-enabled home gateways, and that it planned to
support IPv6 across their entire network by the end of 2012.14
In another similar example, mobile network provider T-Mobile
also announced in April that it had completed the deployment
of IPv6 services across its entire U.S. network, with a published
report noting that it was the largest wireless IPv6 deployment
in the world.15
60005000
40003000200010000
2012201120102009
1-Apr
8-Apr15-Apr22-Apr29-Apr5-May13-May20-May27-May
3-June
10-June17-June24-June
figure 6: Total Number of Autonomous Systems in the IPv6 Routing Table
SECTION 2:
Internet Penetration (continued)
SECTION 2:
Internet Penetration (continued)
2.4 World IPv6 Launch
Building on the success of 2011’s “World IPv6 Day” event, the
Internet Society organized “World IPv6 Launch”, which took
place on June 6, 2012. The intent of this year’s event was to
establish IPv6 as the “new normal” for the Internet, with a list
of participants that include the top Web sites, content delivery
networks, home router manufacturers, and Internet Service
Providers in more than 100 countries around the world.16 With
an IPv6 footprint in over 50 countries at the time of the event,
Akamai supported customers that chose to participate in the
event by enabling them to make content available over IPv6 for
World IPv6 Launch. (This content remains available over IPv6
after the event as well, and new customers have also made
content available via IPv6 through Akamai since the event.)
Figure 7 highlights IPv6 hits/second traffic on the Akamai
Intelligent Platform during the second quarter of 2012, and
is taken from the “Historical” tab of the data visualization at
http://www.akamai.com/ipv6. As is clearly evident within the
graph, IPv6 traffic levels on Akamai grew significantly as several
major customers IPv6-enabled their Web sites and embedded
content. However, because so many customers enabled IPv6
support ahead of World IPv6 Launch, there was no “big spike”
that occurred in the days immediately preceding, or during the
day of, the event. Traffic continued to grow into the World IPv6
Launch event, reaching a quarterly peak several days thereafter,
80000
70000
60000
50000
40000
30000
20000
10000
0
and traffic patterns for the remainder of the quarter appeared
to settle into something of a weekly peak & trough pattern that
is similar to traffic graphs from content delivered over IPv4.
Several weeks after the World IPv6 Launch event, Akamai published
a blog post (“A Data-Driven View of IPv6 Adoption” at https:/
blogs.akamai.com/2012/07/a-data-driven-view-of-ipv6-adoption.
html)that included the infographic shown in Figure 8. Both the
blog post and infographic highlight the year-over-year growth
that Akamai observed in IPv6 traffic, including:
.
A 67x increase in the number of unique IPv6 addresses
making requests for content
.
A 460x increase in the number of requests made for content
over IPv6
.
A 9x increase in requests from end users in the United States
made against a dual-stack (IPv4 & IPv6) consumer-oriented
Web site
The infographic also calls out that nearly three-quarters of the
native IPv6 addresses observed were from the United States,
and that over one-third of the requests were from Verizon Wireless,
a mobile network provider in the United States. This latter
observation is not surprising, given that the Verizon Wireless 4G
LTE network has IPv6 built-in, with the requirement that all LTE
devices must be IPv6 capable.17
World IPv6 Launch
June 6, 2012
1-Apr
8-Apr15-Apr22-Apr29-Apr5-May13-May20-May27-May
3-June
10-June17-June24-June
figure 7: Akamai IPv6 Hits/Second Traffic Levels, Q2 2012
10
t
Requests
280,229 18,999,253 8,343,590 3,394,971,156
2011
2012
2011
2012
67x 460x
The above comparisons are for 24-hour periods (6/8/2011 and 6/6/2012)
figure 8: World IPv6 Launch Infographic
REQUESTS FROM U.S. END USERS AGAINST
A DUAL-STACK CONSUMER-ORIENTED SITEPREFER IPV6 (%)
AUG
2011
0.0
0.2
0.4
0.6
0.8
1.0
1.2
SEP
2011
OCT
2011
NOV
2011
DEC
2011
JAN
2012
FEB
2012
MAR
2012
APR
2012
MAY
2012
JUN
2012
JUL
2012
For a real-time visualization of IPv6 trafc, please visit www.akamai.com/IPv6
9x increase over the past year
End-user growth in theU.S.
1.04
0.89
0.63
0.59
0.36
0.25
0.190.180.150.12
0.47
0.41
2.5 New Generic Top Level Domains (gTLDs)
Generic top level domains (gTLDs) have been in use on the
Internet since 1985, when IETF RFC 92018 specified the initial
set of five (.gov, .edu, .com, .mil, and .org). Additional gTLDs
including .biz, .info, .name, .museum, .coop, .pro, and .aero
were activated during 2001 . 2002,19 and other gTLDs have
been added over time as well, resulting in a total of 22 gTLDs20
in the domain name system.
The Internet Corporation for Assigned Names and Numbers
(ICANN) voted on June 20, 2011 to end most of the restrictions
on gTLDs from the currently available set of 22, allowing
companies and organizations to choose essentially arbitrary
top-level domain names. (The expectation was that most
applicants would pursue gTLDs associated with their brands.21)
ICANN started to accept applications for new gTLDs on January
12, 2012, with an associated deposit fee of $5,000 USD
and an evaluation fee of $185,000 USD.22 The application
process closed on May 30, and on June 13, “Reveal Day”
occurred, during which ICANN published the list of nearly
2,000 gTLDs that were applied for, including information
on the organizations that submitted the applications.23
According to ICANN,24 applications were received from
60 countries, including:
.
911 from North America.
.
675 from Europe.
.
303 from the Asia-Pacific region.
.
24 from Latin America and the Caribbean.
.
17 from Africa.
In addition, 116 applications were for Internationalized Domain
Names, for strings in scripts such as Arabic, Chinese, and Cyrillic.
Examining the list, one can see that there are quite a few applications
related to brands (like .AMEX, .ALLSTATE, and .ALFAROMEO),
others related to specific geographical locations (like .AFRICA
and .AMSTERDAM), as well as a number that have multiple
applicants competing for the right to a given gTLD (like .APP
and .ART). As expected, Google, Amazon, Microsoft, and
others applied for multiple gTLDs -- some generic terms, others
related to their products/services. However, it will be some time
before any of these new gTLDs are approved and activated for
use on the Internet, as there are associated objection, evaluation,
dispute resolution, and pre-delegation processes/stages
that each application needs to pass through, according
to the new gTLD FAQ published by ICANN.25
World IPv6 Launch
Acloserlook at whaAkamai sawduring
IPv6IPv6
Addresses
Observations
Types of Addresses
16.5
MILLION
2.6
NATIVE IPv6
MILLION
6to4 & TEREDO
Native IPv6 Address
Distribution by Geo
USA73%
FRANCE15%
JAPAN 3.7%
ROMANIA 3%
GERMANY 0.7%
OTHER EUROPE 3%
OTHER ASIA 1.4%
OTHER <1%
Request Distribution
by Network
VERIZON
WIRELESS (US)
38.1%
AT&T (US)
18.1%
FREE (FRANCE)
16.6%
RCS & RDS
(ROMANIA)
7.3%
COMCAST (US)
3.7%
TANet (TAIWAN)
0.9%
KDDI (JAPAN)
0.8%
THOUSANDS
OF OTHERS
14%
11
SECTION 3:
Geography. Global
SECTION 3:
Geography. Global
By virtue of the approximately two trillion requests for Web content that it services on a daily
basis through its globally deployed Intelligent Platform, Akamai has a unique level of visibility
into the speeds of end-user connections and, therefore, into broadband adoption around
the globe. Because Akamai has implemented a distributed platform model, deploying
servers within edge networks, it can deliver content more reliably and consistently than
centralized providers that rely on fewer deployments in large data centers. For more
information on why this is possible, please see Akamai’s How Will The Internet Scale?
white paper26 or the video explanation at http://www.akamai.com/whytheedge.
The data presented within this section was collected during
the second quarter of 2012 through Akamai’s globally-deployed
Intelligent Platform and includes all countries that had more
than 25,000 unique IP addresses make requests for content to
Akamai during the quarter. For purposes of classification within
this report, the “high broadband” data included below is for
connections at greater than 10 Mbps, and “broadband”
is for connections of 4 Mbps or greater. As noted in last quarter’s
report, these definitions have been updated to reflect an overall
trend toward greater availability of higher speed connections.
Similarly, as noted last quarter, the State of the Internet report
will no longer include “narrowband” (connections of 256 kbps
or less) data, nor will it include city-level data.
In addition to providing insight into high broadband and
broadband adoption levels, the report also includes data on
average and average peak connection speeds.the latter provides
insight into the peak speeds that users can likely expect
from their Internet connections.
Finally, traffic from known mobile networks will be
analyzed and reviewed in a separate section of the report;
mobile network data has been removed from the data set
used to calculate the metrics in the present section, as well
as subsequent regional “Geography” sections.
3.1 Global Average Connection Speeds
The global average connection speed once again saw a solid
quarter-over-quarter increase, growing 13% to reach 3.0
Mbps, as shown in Figure 9. However, quarterly changes were
not as positive across the top 10 countries, with eight seeing
lower average connection speeds than in the previous quarter.
Among this group, only Switzerland and the Czech Republic
grew quarter-over-quarter, increasing 4.0% (to 8.4 Mbps) and
0.7% (to 7.2 Mbps) respectively. The quarterly declines seen in
the other eight countries were fairly nominal, ranging from
a loss of just half a percent in Denmark (to 6.7 Mbps) to a loss
of just under 10% in South Korea (to 14.2 Mbps). Globally, 69
total countries that qualified for inclusion saw average connec
figure 9: Average Measured Connection Speed by Country
. Global 3.0 13% 15%
1 South Korea 14.2 -9.8% 2.9%
2 Japan 10.7 -1.6% 21%
3 Hong Kong 8.9 -4.1% -14%
4 Latvia 8.7 -1.1% 5.5%
5 Switzerland 8.4 4.0% 16%
6 Netherlands 8.0 -9.6% -6.8%
7 Czech Republic 7.2 0.7% -2.5%
8 Denmark 6.7 -0.5% 9.1%
9 United States 6.6 -1.4% 16%
10 Finland 6.6 -4.1% 16%
countryyoy
change
QoQ
change
Q2 ‘12
avg. mbps
country
9
2
3
1
7
7
45
6
10
8
12
tion speeds increase quarter-over-quarter, ranging from a 227%
increase in Kenya (to 1.8 Mbps) to a barely perceptible increase
of just 0.1% in the Ukraine (to 4.4 Mbps). In contrast, 65 total
qualifying countries saw average connection speeds decline
quarter-over-quarter, ranging from losses of just a tenth of a
percent in Lithuania and Germany (to 5.1 Mbps and 5.8 Mbps
respectively) to a 29% decline in Cote D’Ivoire (to just 0.4 Mbps).
Long term trends were generally more positive, with the global
average connection speed growing 15% year-over-year, and
seven of the top 10 countries also seeing increases year-over
year. Four countries saw particularly strong growth, with Japan,
Switzerland, the United States, and Finland all growing more
than 10%. Among the top 10, only Hong Kong, the Netherlands,
and the Czech Republic lost ground year-over-year,
with the 14% decline seen in Hong Kong the most significant.
Globally, 100 qualifying countries saw year-over-year increases,
ranging from 242% in Kenya to just 0.4% in New Zealand (to
3.9 Mbps). Year-over-year declines were seen in 34 countries,
with losses ranging from a meager 0.3% in Bangladesh
(to 0.7 Mbps) to Libya’s 69% decline (to 0.5 Mbps).
According to the Global Information Technology Report of
2012, Kenya was ranked 136th out of 142 countries in the
broadband subcategory.27 However, the changes noted above
do point towards improved connectivity within the country,
possibly related to recently-laid fiber optic cables that aim to
improve broadband interconnectivity among East African
nations. In May, the government of Kenya invited the newly
created nation of South Sudan to connect to these recently-laid
fiber optic cables.28 In addition, submarine telecommunications
network provider Seacom noted in May that it was putting plans
into place to double capacity on its system later in 2012.the
system brings multi-terabit Internet connectivity to Kenya, connecting
it with countries including South Africa and France.29
In the second quarter, 22 qualifying countries had average
connection speeds of 1 Mbps or less, down from 24 in the first
quarter of 2012. Cote D’Ivoire displaced Libya as the country
with the lowest average connection speed, declining 29%
from the prior quarter to 0.4 Mbps, as noted above.
3.2 Global Average Peak Connection Speeds
The average peak connection speed metric represents an average
of the maximum measured connection speeds across all of the
unique IP addresses seen by Akamai from a particular geography.
The average is used in order to mitigate the impact of
unrepresentative maximum measured connection speeds. In
contrast to the average connection speed, the average peak
connection speed metric is more representative of Internet
connection capacity. (This includes the application of so-called
speed boosting technologies that may be implemented within
the network by providers in order to deliver faster download
speeds for some larger files.)
The global average peak connection speed one again showed
strong improvement, growing 19% in the second quarter to
16.1 Mbps. However, Figure 10 shows that this strong improvement
was not reflected in the quarter-over-quarter changes
seen in the top 10 countries. Five countries saw quarterly
13
3
1
2
4
10
7
6
5
8
figure 8: Average Peak Connection Speed by Country
. Global 16.1 19% 44%
1 Hong Kong 49.2 -0.2% 11%
2 South Korea 46.9 -1.9% 31%
3 Japan 40.5 2.4% 28%
4 Romania 38.6 -0.6% 15%
5 Latvia 33.5 . 14%
6 Switzerland 29.9 4.3% 25%
7 Belgium 29.5 1.1% 10%
8 Singapore 28.3 -1.1% 36%
9 Hungary 28.0 0.3% 15%
10 Bulgaria 27.9 1.7% 17%
…
13 United States 27.1 -5.4% 24%
yoy
change
QoQ
change
Q2 ‘12
Peak mbps
country
figure 10: Average Peak Connection Speed by Country
9
13
SECTION 3:
Geography. Global (continued)
SECTION 3:
Geography. Global (continued)
increases in average peak connection speeds, though they
were significantly more modest than the global increase,
with growth ranging from 0.3% in Hungary (to 28.0 Mbps)
to 4.3% in Switzerland (to 29.9 Mbps). Latvia remained flat
quarter-over-quarter, at 33.5 Mbps, while the remaining 4
countries and the United States (at #13) all saw quarterly
declines in average peak connection speeds. Globally, a total
of 54 countries that qualified for inclusion also saw quarterly
declines in average peak connection speeds, with changes
ranging from a loss of just a tenth of a percent in Trinidad
and Tobago (to 17.7 Mbps) to a significantly larger 27% loss
in Oman (to 5.5 Mbps). On a more positive note, 79 qualifying
countries around the world saw average peak connection
speeds increase quarter-over-quarter. The largest increase was
seen in Kenya, which grew 161% to 7.2 Mbps. The smallest
increase was seen in Hungary, as noted above.
Looking at year-over-year changes, significant improvement
was once again seen in the global average peak connection
speed, which grew 44%. Yearly increases were seen across all
of the top 10 countries, as well as in the United States, with
growth of 10% or more seen in all countries. Year-over-year
changes ranged from 10% in Belgium (to 29.5 Mbps) to 36%
in Singapore (to 28.3 Mbps). Globally, 126 qualifying countries
saw year-over-year increases in average peak connection speeds,
including six countries that grew in excess of 100%. Growth
ranged from 1.4% in Luxembourg (to 16.6 Mbps) to 216%
in Kenya. Quarterly declines were seen in only eight qualifying
countries, ranging from a 4.7% loss in Tanzania (to 5.6 Mbps)
to a sizable 54% drop in Oman.
3.3 Global High Broadband Connectivity
As was noted in the 1st Quarter, 2012 State of the Internet
report, the term “high broadband” (as used within the
report) was redefined to include connections to Akamai
of 10 Mbps or greater.
After a sizable increase seen last quarter, the global high broad
band adoption rate declined slightly in the second quarter, losing
1.6%, but remained at 10%. As shown in Figure 11, seven of
the top 10 countries also had negative quarter-over-quarter
changes, with wildly varying magnitudes of change, ranging
from a trivial loss of just 0.6% in Latvia (to 26%) to a much
more concerning decline of 24%, seen in both the Netherlands
and Belgium (to 17% and 14% respectively). Of the three
countries that saw high broadband adoption levels improve
quarter-over-quarter, the United States grew 5.5% (to 16%),
while Switzerland and the Czech Republic saw solid growth,
increasing 15% (to 22%) and 21% (to 14%) respectively. On
a global basis, 20 countries that qualified for inclusion saw high
broadband adoption levels increase quarter-over-quarter, rang
ing from Italy’s 1.6% increase (to 2.6%) to an extremely strong
56% increase in South Africa (to 1.1%). Among the 25 qualify
ing countries that saw high broadband adoption levels decline
quarter-over-quarter, losses ranged from just half a percent in
Germany (to 8.3%) up to 35% in Portugal (to 4.4%). In the
second quarter, all of the top 10 countries had high broadband
adoption levels above the global average of 10%, as did six
additional countries. China continued to have the lowest level
of high broadband adoption, giving back the increase seen
in the first quarter, but remaining at just 0.1%.
7
410
2
3
1
9
5
6 8
. Global 10% -1.6% 25%
1 South Korea 49% -7.4% 49%
2 Japan 37% -1.8% 22%
3 Hong Kong 26% -6.9% -12%
4 Latvia 26% -0.6% 7.1%
5 Switzerland 22% 15% 59%
6 Netherlands 18% -24% -16%
7 United States 16% 5.5% 76%
8 Czech Republic 14% 21% 19%
9 Belgium 14% -24% 25%
10 Finland 14% -16% 46%
yoy
change
QoQ
change
% above
10 mbps
country
figure 11: High Broadband (>10 Mbps) Connectivity
14
Looking at year-over-year changes, the global high broadband
adoption level saw a solid 25% increase as compared to the
second quarter of 2011. Among the top 10 countries, none
saw adoption levels double year-over-year (as three did last
quarter), though Switzerland and the United States both posted
increases of more than 50%. Five other countries among the
top 10 experienced double-digit yearly growth rates, while Latvia
was the only country among the group that grew less than 10%
year-over-year, with a still respectable 7.1% increase (to 26%).
Around the world, a total of 34 countries that qualified for inclusion
saw high broadband adoption levels increase year-over-year,
with five countries seeing adoption levels more than double.South
Africa’s 284% was far and away the largest increase. The smallest
year-over-year increase was seen in the Ukraine, which added
2.6% (to 6.3%). Eleven qualifying countries saw high broadband
adoption levels decline year-over-year, ranging from a 0.3%
decline in Slovenia (to 5.4%) to 27% losses in both Turkey and
Portugal (to 0.5% and 4.4% respectively).
3.4 Global Broadband Connectivity
As was noted in the 1st Quarter, 2012 State of the Internet report,
the term “broadband” (as used within the report) was redefined
to include connections to Akamai of 4 Mbps or greater.
After a solid increase in the first quarter, the global broadband
adoption level saw a minor decrease in the second quarter, losing
2.8% and declining to 39%. Six countries among the top 10, as
well as the United States, also saw quarterly declines in broadband
adoption levels. As shown in Figure 12, these quarterly declines
were fairly nominal, with the largest seen in the Czech Republic,
which lost 6.4%, (to 64%). Quarterly increases across the remain
ing four countries in the top 10 were also fairly nominal, ranging
from just 0.8% in Canada (to 69%) to 5.2% in Denmark (to 66%).
Globally, just 25 countries that qualified for inclusion saw higher
broadband levels than in the prior quarter, with adoption levels
more than doubling in both Kenya and Morocco. Fifty-five coun
tries around the world had broadband adoption levels greater than
10%, while India and Indonesia had the lowest levels of adoption
among countries on the list, at 1.4% and 0.8% respectively.
Looking at year-over-year changes, global broadband adoption
increased 6.9%, while increases were also seen in just half of
the top 10 countries, as well as the United States. Among the
five countries that saw yearly growth, increases ranged from
5.0% in Denmark to 28% in South Korea (to 84%). Yearly declines
among the other five countries were fairly modest, with
the largest seen in the Czech Republic, which lost 10%. Glob
ally, 46 total countries that qualified for inclusion saw broad
band adoption levels increase year-over-year, with growth above
100% seen in nine countries, and another 29 seeing growth of
10% or more. Lithuania had the lowest level of yearly growth,
increasing just 0.8% to 51% adoption. Among the qualifying
countries where broadband adoption levels declined year-over
year, losses ranged from a meager 0.1% in Cyprus (to 9.4%) to
a surprisingly large 47% drop in Vietnam (to 3.0%).
figure 12: Broadband (>4 Mbps) Connectivity
. Global 39% -2.8% 6.9%
1 South Korea 84% -2.2% 28%
2 Switzerland 79% 2.3% 14%
3 Netherlands 79% -5.0% -2.0%
4 Japan 74% 2.4% 20%
5 Belgium 69% -4.7% -6.3%
6 Canada 69% 0.8% 8.2%
7 Hong Kong 68% -5.2% -4.7%
8 Latvia 67% -1.7% -2.8%
9 Denmark 66% 5.2% 5.0%
10 Czech Republic 64% -6.4% -10%
…
13 United States 57% -4.6% 7.9%
yoy
change
QoQ
change
% above
4 mbps
country
6
13
4
7
1
9
3
5
810
2
15
SECTION 4:
Geography. United States
SECTION 4:
Geography. United States
The metrics presented here for the United States are based on a subset of data used
for Section 3 and are subject to the same thresholds and filters discussed within that
section. (The subset used for this section includes connections identified as coming
from networks in the United States, based on classification by Akamai’s EdgeScape
geolocation tool.) As was noted in the introduction to Section 3, this section no longer
includes city-level data nor data on narrowband (<256 kbps) connections, and the
“new” definitions of high broadband (>10 Mbps) and broadband (>4 Mbps), put into
place starting with last quarter’s report, are used here as well.
4.1 United States Average Connection Speeds
Consistent with its standing in the prior quarters, Delaware
continued to rank as the fastest state in the union in the second
quarter of 2012, improving its average connection speed to
12.1 Mbps, an 18% quarter-over-quarter increase. As shown in
Figure 13, nine of the top 10 state saw positive quarter-overquarter
changes in average connection speeds, with the largest
increase seen in Delaware, and the smallest, at 3.8%, seen in
Connecticut. The 8.2% increase in New Hampshire pushed the
state’s average connection speed past the 10 Mbps mark,
allowing it join Delaware in that respect.Across the whole country,
39 total states saw average connection speeds increase in the
second quarter, with six of those states growing in excess of
10%. The largest increase was seen in New Mexico, at 19%
(to 6.0 Mbps), while the smallest increase was seen in Nebraska,
at 0.1% (to 5.8 Mbps). A 12% quarter-over-quarter decline
in Missouri allowed it to displace Arkansas as the state with
the lowest average connection speed, at 3.6 Mbps.
Looking at year-over-year trends, all of the top 10 states saw
average connection speeds increase, with fairly significant
growth seen across the states on the list.the smallest increase
was seen in Rhode Island, at just over 9%, while New Hampshire’s
54% increase was the largest. Across the whole country,
45 states and the District of Columbia saw increased average
connection speeds as compared to the second quarter of 2011,
and over 30 states saw double digit percentage increases.
Several announcements made in the second quarter point
toward continued improvements in connection speeds across
the United States. Gig.U is a group of over 30 research universities
from across the United States that aims to bring high speed
broadband connectivity to the communities surrounding these
universities via a public-private partnership model. To date,
Gig.U has worked with private companies to build projects in
Maine and Atlanta, and in May, it announced that it will be
working with startup Gigabit Squared, to select and deploy
yoy
change
QoQ
change
Q2 ‘12
avg. mbps
State
figure 13: Average Measured Connection Speed by State
1 Delaware 12.1 18% 49%
2 New Hampshire 10.1 8.2% 54%
3 District Of Columbia 9.7 9.3% 31%
4 Vermont 9.7 6.9% 38%
5 Rhode Island 9.0 5.4% 9.2%
6 Massachusetts 8.8 7.3% 35%
7 Connecticut 8.7 3.8% 34%
8 Virginia 8.3 6.0% 22%
9 Washington 8.3 5.2% 26%
10 Utah 8.1 -5.7% 15%
1
5
7
8 3
6
4 2
10
9
16
1 Delaware 41.6 -4.1% 35%
2 District Of Columbia 37.5 8.8% 35%
3 Vermont 35.1 -0.7% 37%
4 New Hampshire 34.2 1.3% 42%
5 Virginia 34.1 0.6% 28%
6 Rhode Island 33.6 0.3% 7.5%
7 Washington 32.7 3.7% 40%
8 Massachusetts 32.4 -1.3% 36%
9 New York 31.3 -3.6% 24%
10 Connecticut 31.3 -0.3% 40%
yoy
change
QoQ
change
Q2 ‘12
Peak mbps
State
10
3
8
47
1
6
25
9
figure 14: Average Peak Connection Speed by State
fast broadband to six more communities.30 While the names
of the six communities were not named within the announcement,
the small town of Orono, Maine, announced a plan to
build a local gigabit network, in conjunction with the University
of Maine and local network provider GWI . Gig.U executive
director Blair Levin noted that “the University of Maine was one
of the first institutions to sign-up to participate in Gig.U and is
now one of the first to move forward, in partnership with GWI,
to make the idea of Gig.U a reality…”31
In addition, in June, the White House announced32 the signing
of an Executive Order by President Obama that is intended “to
make broadband construction along Federal roadways and
properties up to 90 percent cheaper and more efficient.” According
to the announcement, “The new Executive Order will ensure
that agencies charged with managing Federal properties and
roads take specific steps to adopt a uniform approach for allowing
broadband carriers to build networks on and through those
assets and speed the delivery of connectivity to communities,
businesses, and schools.” The announcement also highlighted
that nearly 100 partners . including more than 25 cities as
well as corporate and non-profit entities . would be joining
with more than 60 national research universities to form a new
public-private partnership called “US Ignite” to create a new
wave of services that “take advantage of state-of-the-art, programmable
broadband networks running up to 100 times faster
than today’s Internet.”
4.2 United States Average Peak Connection Speeds
Even with a slight quarter-over-quarter decline to 41.6 Mbps,
Delaware continued to have the highest average peak connection
speed in the second quarter. As shown in Figure 14, a decline
in Vermont’s average peak connection speed pushed it down to
third place, allowing it to be bested by the District of Columbia,
which grew 8.8% from the first quarter (to 37.5 Mbps). Four
other states in the top 10 also saw average peak connection
speeds increase quarter-over-quarter, while Massachusetts, New
York, and Connecticut joined Delaware and Vermont in experiencing
quarterly declines. Across the whole country, 21 states
and the District of Columbia grew average peak connection
speeds quarter-over-quarter .New Mexico’s 11% increase (to
25.7 Mbps) was the largest seen. Interestingly, New Mexico had
the lowest level of quarterly growth in the first quarter, though
it was still fairly high, at 8.1%. In contrast, this quarter, eight
states grew less than one percent quarter-over-quarter, with
Florida and Ohio both adding just one tenth of a percent (to
29.5 Mbps and 26.3 Mbps respectively). Similar to the average
connection speed metric, Missouri’s 14% quarterly decline (to
15.7 Mbps) made it the state with the lowest average peak
connection speed . former last place state Arkansas grew
7.0% to 17.7 Mbps.
Year-over-year changes among the top 10 states were all positive,
and in general, fairly major. Rhode Island, which had yearly
growth of 7.5%, was the only state among the group to grow
less than 10%. Among the other states in the top 10, three
grew 40% or more, another four grew in excess of 30%, and
the remaining two grew more than 20%. Across the whole
country, all of the states in the U.S. saw yearly improvements
in average peak connection speeds, with all but five states
growing 10% or more. Particularly strong growth was seen
in Oregon and Montana, which both increased 49% year-overyear
(to 28.6 Mbps and 21.6 Mbps respectively), as well
as in Alaska, which increased 59% (to 24.5 Mbps).
17
SECTION 4:
Geography. United States (continued)
1 Delaware 39% 19% 153%
2 New Hampshire 33% 14% 298%
3 District Of Columbia 32% 17% 91%
4 Vermont 31% 17% 196%
5 Massachusetts 29% 22% 179%
6 Rhode Island 29% 18% 39%
7 Washington 24% 14% 100%
8 Connecticut 24% 9.1% 203%
9 Pennsylvania 23% 22% 169%
10 New Jersey 23% . 160%
yoy
change
QoQ
change
% above
10 mbps
State
8
6
2
1
7
3
109
4
5
figure 15: High Broadband (>10 Mbps) Connectivity, U.S. States
4.3 United States High Broadband Connectivity
As was noted previously, starting with the 1st Quarter, 2012
State of the Internet report, the term “high broadband”, as
used within the report, has been redefined to include connections
to Akamai of 10 Mbps or greater.
As shown in Figure 15, quarter-over-quarter changes among
the top 10 states were generally positive in the second quarter,
with eight of the ten turning in double-digit percentage growth,
Connecticut coming in just below that at 9.1% growth, and
New Jersey remaining unchanged from the first quarter. The
largest quarterly growth among the group was seen in
Massachusetts and Pennsylvania, which both added 22% (to
29% and 23% adoption respectively). Across the whole country,
a total of 37 states and the District of Columbia saw high
broadband adoption levels increase quarter-over-quarter. The
largest increase was seen in New Mexico, which grew 52% (to
12%), while the smallest increase was just a tenth of a percent,
seen in Maine (also to 12%). Arkansas remained the state with
the lowest level of high broadband adoption, though it increased
12% to 3.0%
The levels of year-over-year change in high broadband adoption
rates across the top 10 states were once again extremely significant,
with levels more than doubling in eight states. Of those eight,
New Hampshire and Connecticut saw year-over-year changes in
excess of 200%. Across the remainder of the country, South
Dakota and Kansas also increased more than 200% year-overyear,
while seven other states also saw broadband adoption up
by 100% or more. Nebraska was the only state that grew less
than 10% year-over-year, up 3.0% (to 9.2%), while only Idaho
posted a yearly decline, dropping 3.3% (to 6.0%).
DID YOU KNOW?
.
Across the whole country, a total of 37
states and the District of Columbia saw high
broadband adoption levels increase quarterover-
quarter.
. The levels of ye