Documentos de Académico
Documentos de Profesional
Documentos de Cultura
Exchange
Points
A
closer
look
at
the
differences
between
continental
Europe
and
the
rest
of
the
world.
Table
of
Contents
Abstract
.................................................................................................................................
4
1
Euro-IX
............................................................................................................................
5
2
The
research
.................................................................................................................
6
3
Transit,
peering
and
IXPs
.........................................................................................
7
3.1
How
the
Internet
works
.................................................................................................
7
3.2
Autonomous
Systems
(AS)
............................................................................................
7
3.3
Tier
hierarchy
...................................................................................................................
9
3.4
Transit
vs.
peering
........................................................................................................
10
3.5
What
is
an
IXP?
..............................................................................................................
10
3.5.1
The
purpose
of
an
IXP
..........................................................................................................
11
3.5.2
Connecting
to
an
IXP
............................................................................................................
12
4
A
brief
history
of
IXPs
.............................................................................................
13
4.1
Europe
..............................................................................................................................
13
4.1.1
Western
Europe
......................................................................................................................
13
4.1.2
Northern
Europe
....................................................................................................................
14
4.1.3
Central
Europe
........................................................................................................................
14
4.1.4
Southern
Europe
....................................................................................................................
15
4.1.5
Eastern
Europe
.......................................................................................................................
15
4.2
North
America
................................................................................................................
16
4.3
Asia
....................................................................................................................................
17
4.3.1
China
............................................................................................................................................
17
4.3.2
Japan
............................................................................................................................................
19
4.3.3
India
.............................................................................................................................................
19
4.3.4
Nepal
............................................................................................................................................
20
4.4
Latin
America
.................................................................................................................
20
4.4.1
Brazil
...........................................................................................................................................
20
4.4.2
Argentina
...................................................................................................................................
21
4.5
Africa
.................................................................................................................................
22
5
IXP
models
..................................................................................................................
24
5.1
Commercial
vs.
non-commercial
.............................................................................
24
5.1.1
Neutrality
..................................................................................................................................
24
5.1.2
Ownership
and
management
............................................................................................
24
5.1.3
Fees
..............................................................................................................................................
25
5.1.4
Scope
of
activities
..................................................................................................................
25
5.1.5
Peering
agreements
and
policies
....................................................................................
26
5.2
Europe
..............................................................................................................................
29
5.3
North
America
................................................................................................................
29
5.4
Asia
....................................................................................................................................
30
5.5
Latin
America
.................................................................................................................
30
5.6
Africa
.................................................................................................................................
31
6
Current
developments
and
trends
.....................................................................
32
6.1
IXP
participants
Point
of
View
..................................................................................
32
6.2
Trends
...............................................................................................................................
33
6.2.1
Japan
............................................................................................................................................
34
7
Differences
.................................................................................................................
35
7.1
Connected
parties
.........................................................................................................
35
7.2
Influence
incumbents
..................................................................................................
36
7.3
North
American
vs.
European
IXPs
.........................................................................
36
Abstract
This document describes the research that has been done between August 2010 and January 2011 for the European Internet Exchange Point association. Euro-IX gathers resources about the European IXP scene to be able to make future predictions and improve the IXP community. The research started at looking into the differences between IXPs in the US and Europe, however this soon developed into looking at the differences between IXPs around the world, as it was thought that the initial research would be either very limited or had to be too thorough for the available time. By looking at the entire world a better understanding of the global IXP scene would be gained which would also be of use for Euro-IX, since they have members from outside of Europe. Proper knowledge was needed on the purpose and workings of IXPs and this was gained rapidly. After this a brief history on IXPs in all continents was made to understand their history and current situation a little. After this the differences between IXPs were mapped and located in each continent. As a follow up current trends and developments were looked into, which led to being able to make some future predictions. After doing the research, it can be said that its very difficult to compare the different IXPs to each other, since they are simply too different. Over the next couple of years we can see more IXPs being established in Latin America, Africa and (parts of) Asia, where currently there are very few IXPs present. These IXPs will probably be of a non-profit nature, while in the future more commercial IXPs might also show up. In North America commercial IXPs are dominant, while in Europe there are more non-profit IXPs but the number of commercial IXPs is increasing and is likely to continue to expand over the next couple of years. In North America the connection between networks are mainly private interconnections, while in Europe public peering is more common. European IXPs are likely to offer more services towards their participants, which might also be the case for the now less-developed regions. More peering relationships will be established with the Asian IXP scene, though this will probably be a slow development due to the high transit prices in Asia. It can be said with some certainty, that a large increase in exchanged bandwidth will be seen over the next couple of years, which will lead to a need of IXP platform upgrading to keep up with this demand.
1 Euro-IX
In May 2001 European Internet Exchange Points (IXP) in Europe decided there was need to combine their resources to strengthen, develop and improve the IXP community. This led to the formation of the European Internet Exchange Association (Euro-IX). This development and improvement of the IXP community is achieved by the coordination of technical standards, the development of common procedures, sharing ideas, knowledge and sharing public statistics and information between IXPs. To help accomplish this Euro-IX organises biannual two-day long meetings that all Euro-IX members can attend. They also provide mailing lists and on-line resources. Euro-IX currently has 56 affiliated IXPs from 37 countries around the world and is located in Amsterdam, The Netherlands. There is interest in having a clearer understanding of the differences between the continents as this can help give a better overview of the current situation of IXPs and developments around the world. One of the questions that have been raised for some time now, is why there is a large difference between the amount of unique ASNs at IXPs in Europe compared to the amount connected to North American IXPs, while the total number of advertised ASNs on both continents is quite similar. This is one of the questions that have been examined in this research and an attempt to answer this question will be made.
2 The research
The European Internet Exchange Association (Euro-IX) has been gathering data about IXPs, the parties connected to them and infrastructure to exchange the traffic for almost 10 years now. This has allowed them to produce reports on the current developments and trends, which they now want to use to produce more insights into trends that have been occurring and work on future predictions. Euro-IX is interesting in usage trends that are occurring in other regions than Europe and how these compare to Europe. What the IXPs look like in the US, if there are differences in the way that Internet traffic is handled, if Asian ISPs are as likely to make use of IXPs as those in Europe and if future predictions in different regions look similar. Euro-IX hopes to gain a better picture of the differences between the European and US peering ecosystems, what has and still does influence the amount of organizations that are present at public peering points on both continents, what makes European IXPs, relatively speaking more popular than their counterparts in the US? With a better understanding of these peering ecosystems we hope to better understand and predict the future of the global peering scene. It has to be delivered in a document describing the results and outcomes. Euro-IX wishes to have better insights in trends occurring around IXPs, not only in Europe but also in the US. To be able to do this, better insights are needed of the past and current situation on this continent, what differences there are and how they compare to each other. It is assumed that the main difference in IXP ecosystems between Europe and the US is that in the US major peering points are in fact run by colocation operators with the goal of making a profit via offering collocation space rather than simply serving the community and this has a direct affect on the amount of ASNs seen in the US.
Internet is available all over the world, but how do you reach a website that is located on a different continent? Where does the traffic go when you make an information request and how does it reach its destination? All of this is done by Internet Service Providers (ISPs) who either peer or buy transit and is assisted by Internet Exchange Points (IXPs). This chapter explains the basics of how Internet traffic is transported over the world.
The
definition
of
The
Internet
from
Dictionary.com
[1]: a
vast
computer
network
linking
smaller
computer
networks
worldwide
(usually
prec.
by
the
).
The
Internet
includes
commercial,
educational,
governmental,
and
other
networks,
all
of
which
use
the
same
set
of
communications
protocols.
When you are sitting behind your computer and are connected to the Internet, you can type in a URL in your web browser to request a webpage, hit enter and almost instantly the requested webpage appears on your screen! Or when you message a friend with a chat client, your message appears almost immediately on his screen and he can even message you back. Whats happening behind the scenes is what we call the magic of the Internet. The moment you hit enter, your request or the message you want to send, is transported to your ISP. This is the company that you buy your Internet access from. As a customer of an ISP, you are part of the ISPs network. When you request a webpage, it will be transported to your ISP, from which it will be transported to other networks (The Internet). Before your request reaches its destination it probably went through a lot of different networks (depending on where your request is destined). All of these networks that your request may travel through are called Autonomous Systems. One Autonomous System may be made up of smaller networks, but this is not always the case. All of these Autonomous Systems connecting together make the Internet, as we know it today.
identify the network. In general ISPs and content providers have their own unique ASN, which is used to identify their network to setup BGP [2] sessions. This makes it possible for other networks to connect to them and exchange traffic. Having an ASN is also a requirement of connecting to an Internet Exchange Point (IXP). Figure 2 is an example of how different ASes are connected to one IXP with peering relationships to another AS. The rules of who to peer with, using the IXP, really depends on the IXPs peering rules. Some want all their participants to peer with everybody, while others leave the option of who to peer with up to the participants themselves. Setting up a (peering) relationship with another ISP (or content provider) is usually entered into to decrease the costs for interconnection. Furthermore, the number of AS hops that are required for the traffic to traverse and reach its destination can be reduced. In general IXPs dont interfere with the relationships that are created between their participants.
10
thing: a physical point where different networks can exchange traffic with each other. In Figure 4 you can see different parties connected to a switch at an IXP. When the IXP is non-profit the parties connected to the IXP are generally referred to as members. However, when the IXP is commercial, its participants are referred to as customers. In this paper we refer to the parties connected to the IXP as participants, irrespective of whether the IXP is non-profit or commercial.
3.5.1 The purpose of an IXP The idea of an IXP was started as a way of keeping local Internet traffic local. This means, not using international (transit) lines to reach a destination in the same region as where the request originally came from. IXPs were introduced so that all parties willing to participate could connect at one (local) point, exchange traffic, save money and decrease Round Trip Times (RTT). Using an IXP also resulted in better network performance and QoS (Quality of Service). International lines are always available to use in case an IXP happens to fail, so the end-user wont notice much in terms of performance should this occur. When customers of an ISP enjoy their Internet experience, they are generally willing to pay more for their connectivity and speed. Money paid by customers can be invested in the infrastructure and services to the customers and not simply in international transit costs. This is good for the local economy. With fewer costs
11
for international traffic (because almost all the traffic can be exchanged via the IXP), the prices for consumer internet can go down, which will lead to more customers for the ISP. When a good local Internet infrastructure is available, other businesses and organizations are more willing to establish themselves in the same region, which leads yet again to more customers and also more content for the ISP. 3.5.2 Connecting to an IXP An IXP may have certain requirements that potential participants must meet before they are able to connect to their network. One very common requirement is that the party wishing to connect needs to have an ASN. What the specific joining requirements are really depends on the individual IXP. While their main purpose might be the same, their business models may differ and this may have an influence on your decision to join an IXP or not. Besides that you may also need to have an idea of the amount of bandwidth you plan to exchange on your port at the IXP; if you have any requirements on your housing location its a good idea to look into this as well, since most IXPs have their infrastructure at spread over multiple datacenters.
12
In this section a brief introduction will be given to cover the history and the present day situation of IXPS in regions around the world. Many of the first Internet connections in Europe were initiated by academic- and research-networks [4] or by telecom operators. Together they gathered to create one local point of interconnection, which resulted in IXPs. In the early days of Internet Exchanges in Europe, there was less competition between ISPs, because the biggest telecommunications operator (Telcos) kept the costs of transport, both for transit as for transatlantic, very high. Most of the traffic around this time was destined for America, because all content was available there. The local ISPs were forced to work together to reduce the costs for transporting traffic to North America. After a few years more content became available in Europe, because there was more local interest and this started to play a part in the development. The availability of local content changed the traffic flow. More traffic stayed within the European borders and less traffic went overseas, this most likely played an important role in forming the interconnection landscape. The very first European IXPs were established in 1993. This was around the time when Internet access became more readily available to the home users and more traffic started being exchanged. More traffic had a local destination, which resulted in the need for a local exchange. Peering helped to keep the end-users costs low, the transit and transport costs were high while at the same time the ISPs stayed competitive towards North American based providers. Today, the Internet penetration [5] in Europe is 58.4%, and makes up 24.2% of the Internet users worldwide. 4.1.1 Western Europe In this document we define Western Europe as: Belgium, France, the Netherlands, Luxembourg and the United Kingdom. The very first Internet Exchanges in Western Europe were established in 1994. These IXPs were named LINX (London Internet Exchange) and AMS-IX (Amsterdam Internet Exchange). Today, they both have more than 300 connected members from diverse locations around the world, with traffic peaks of over 1.1Tbps at AMS-IX [6] and more than 700Gbps at LINX [7]. There are approximately 30 IXPs spread across Western Europe. When looking at the IXPs in France, you can see that almost all their IXPs are located in and around Paris, with the recent creation of France-IX supported by Akamai, Jaguar Network, Neo Telecoms and Interxion in the hope of gaining better connectivity around France. The Internet penetration [5] in Western Europe is 2nd highest in Europe, United Kingdom has 82.5%, the Netherlands with 88.6%, Belgium 77.8%, Luxembourg 85.3% and France 68.9%.
13
4.1.2 Northern Europe In this document we define Northern Europe as: Denmark, Finland, Iceland, Norway and Sweden Northern Europe is the most progressive region of Europe based on Internet access and technology. The first IXPs in Northern Europe were established in 1993 in Finland, going by the name of FICIX the Finish Commercial Internet Exchange, created by the Finish telecoms and NIX, the Norwegian Internet Exchange in Oslo, established by the University of Oslo. In 1994 the Danish Internet Exchange (DIX) was established in Copenhagen and a few years later, in 1996 two more IXPs were established Netnod in Sweden and NorrNod in Norway. After this the development of IXPs the Internet connectivity grew tremendously. Together with Western Europe, Northern Europe has always been one of the leading regions in Europe as far as Internet connectivity goes. The Internet penetration [5] in Northern Europe is the highest in Europe with a penetration of 86.1% in Denmark, 85.3% in Finland, 94.8% in Finland, 92.5% in Sweden and 97.6% in Iceland. This makes Northern Europe possibly the best- connected region in Europe with Western Europe following close behind. Figure 5.1 shows all European countries that have one or multiple IXPs. The countries in grey are those that do not yet have an IXP. The white countries are outside of European borders. 4.1.3 Central Europe In this document we define Central Europe as: Austria, Czech Republic, Germany, Hungary, Liechtenstein, Poland, Slovakia, Slovenia and Switzerland. In 1995 one of the todays largest IXPs was established in Frankfurt, Germany known as DE-CIX [8] (Deutsche Commercial Internet Exchange). They currently have more than 300 customers with IXP traffic exceeding 1.2Tbps. Germany has an Internet penetration of 79.1%. In other Central European countries many smaller to mid-sized IXPs can be found. After DE-CIX, the largest IXPs in Central Europe are NIX.CZ in Prague, Czech Republic and PLIX in Warsaw, Poland both with traffic peaks in excess of 100Gbps. In Central Europe the Internet
14
penetrations [4] is between 58.4% and 75.3% (Germany not included). Its interesting to see that the country with the lowest Internet penetration (Poland with just 58.4%) has one of the biggest traffic exchanging IXPs in Central Europe, while Slovakia; with their IXP exchanging less traffic (about 40Gbps) has an Internet penetration of 74.3%. 4.1.4 Southern Europe In this document we define Southern Europe as: Albania, Bosnia and Herzegovina, Bulgaria, Croatia, Greece, Macedonian, Italy, Portugal, Serbia and Spain. Most of the Southern European IXPs were established around the late 1990s. In 1997 ESPANIX (Spains Internet Neutral Exchange Point) was established in Madrid connecting the most important Spanish telecom operators and are currently one of the largest IXPs in the Southern European region. Other countries with notable amounts of traffic at their IXPs are Bulgaria, Italy and Greece. The Internet penetration [4] in Southern Europe is significantly lower compared to that of the Northern and Western European regions, here it lies between 40% and 65%, while Bosnia Herzegovina at just 31.2%. 4.1.5 Eastern Europe In this document we define Eastern Europe as: Belarus, Estonia, Latvia, Lithuania, Moldova, Romania, Russia and Ukraine. Most the IXPs in Eastern Europe were established around 2000 and 2001. When looking at all the countries in this region, you can see that the traffic exchanged in Ukraine and Russia are tremendously high compared to the other countries, where most of the time the traffic is between 1Gbps and 10Gbps. The UA-IX (Ukrainian Internet Exchange) exchanges about 220Gbps at its peak and the MSK-IX (Moscow Internet Exchange) has peak traffic of around 400Gbps. Its interesting to see that quite a few Eastern European telecom operators and businesses connect to DE-CIX in Germany to get more routes to the rest of the Internet. This is probably due to Germanys geographical proximity to the Eastern European region. The Internet penetration [4] in Eastern Europe is comparatively lower to that of other European regions, however its growing rapidly, thus the penetration will potentially be even higher in the next couple of years. Estonia, Latvia and Lithuania have the highest Internet penetration with 75.1%, 67.8% and 59.3% respectively. In Belarus this figure is 46.2%, Moldavia 30.0% and Romania 35.5%. Its interesting to note that the Internet penetration is somewhat lower in Russia with just 42.8% and Ukraine with 33.7%, while the amount traffic at their IXPs is substantially higher than other countries in not only the region but the whole of Europe.
15
16
connect to the network. This became a burden for the NSFNET, as there were too many people while not enough capacity to handle all this traffic. To ensure connectivity to all connected companies and organizations the National Science Foundation awarded contracts to four Network Access Points (NAP). Different companies spread across North America ran all these NAPs. The companies that received these contracts were: The Pacific Bell NAP in San Jose (California), the Ameritech NAP in Chicago (Illinois), the Sprint NAP in Pennsauken (New Jersey) and the already existing MAE East, operated by MCI WorldCom in Vienna (Virginia). In the coming years the amount of traffic started to reduce, because larger networks decided to set up a private interconnection and the smaller networks were limited to regional public interconnections Today, approximately 90 Internet Exchanges are located in North America and most of them are operated as a commercial service. These commercial Internet Exchanges havent been confined to within the American borders, but have reached out to Europe and Asia as well. Non-commercial Internet Exchanges can also be found in North America, however numbers are far less. How much IXP traffic is being exchanged within this region is a little unclear, as many of the IXPs do not publicly display these statistics. About 5% of the worlds population is located in North America with an Internet penetration [5] of over 75%. The US makes up about 13% of the total number of Internet users in the world.
4.3 Asia
There
is
a
large
difference
between
the
countries
within
Asia
with
their
level
of
Internet
access
and
connectivity.
For
instance
countries
like
China
and
Japan
are
more
developed
than
countries
like
Vietnam,
Indonesia,
Nepal
and
India,
where
the
Internet
penetration
is
significantly
lower.
When
you
look
at
the
population
statistics
[5]
from
Asia,
you
can
see
that
56.0%
of
the
world
population
is
located
in
Asia.
Of
the
total
Asian
population,
only
21.5%
has
access
to
Internet.
In
the
last
10
years
the
Internet
growth
has
increased
tremendously,
and
this
seems
to
be
a
continuing
trend.
About
42.0%
of
the
total
Internet
Users
in
the
world
are
located
in
Asia.
This
would
suggest
that
a
lot
of
traffic
is
going
through
the
Asian
regions.
4.3.1 China
One
of
the
very
first
Internet
links
(64
kbps)
in
Hong
Kong
was
in
1991
by
the
Chinese
University
of
Hong
Kong
(CUHK)
[10].
However,
back
in
those
days
it
was
still
under
British
rule.
A
year
later
all
higher
educational
institutions
in
Hong
Kong
were
linked
to
each
other
over
this
link.
They
called
this
network
HARNET
(Hong
Kong
Academic
and
Research
NETwork).
In
1994
when
a
few
small
ISPs
were
established
the
Internet
on
the
non- academic
side
started
to
grow
while
serving
mostly
individual
subscribers.
The
real
growth
of
Internet
started
in
1995
as
more
and
more
companies
started
to
see
the
importance
of
the
Internet.
This
caused
a
major
impact
on
the
availability
of
local
content
and
intra-Hong
Kong
communication
became
more
of
a
necessity.
However
a
large
number
of
ISPs
still
had
to
rely
on
other
ISPs
that
had
a
local
presence
to
complete
there
do
transit
for
them.
This
massive
growth
of
Internet
caused
a
need
for
a
better
network
infrastructure
so
as
to
lower
the
costs
of
operations.
It
was
not
very
effective
for
ISPs
to
route
their
intra-Hong
Kong
traffic
overseas,
because
the
links
were
17
expensive and relatively slow, thus CUHK set up the framework for HKIX and by negotiating with newly established ISPs with links to overseas and agreed on getting a link to CUHK. The ISPs placed their own routers at CUHK, while they provided the space, electricity, and a network to connect all routers of the participants. The philosophy of HKIX is that every participant is equal and will be treated fairly and together make certain that traffic is routed over the fastest possible path and all participants can gain the greatest possible benefits. The growth of HKIX has led to having almost 90 licensed and over 40 non- licensed members with aggregated traffic peaks from over 120Gbps and still growing The largest network provider in China is ChinaNet, owned by China Telecom. They carry about 80% of traffic from the Chinese mainland, excluding traffic from Hong Kong. In 2000 Chinas first NAP was opened: Beijings National Internet Exchange Centre and connects all major network providers. In the same year the Shanghai NAP was established, mainly to connect major ISPs in Shanghai and its surrounding districts. The Internet penetration [5] within China is about 31.6%, this is about 50.9% of all Asian Internet users and is the leading country in Asia as far as the number of users is concerned.
Figure 5.3 shows the Asian countries that have one or more IXPs. The white countries have no known IXPs at this moment.
18
4.3.2 Japan In 1975 an academic project [11] was initiated by the Japanese Government, which was developed to connect Tokyo University and Kyoto University. Which were located 500km apart from each other. This network was given the name N1 Network. In 1981 the network between the universities was officially in use. However the N1 Network had certain limitations, such as: e-mail support, it couldnt interconnect with networks outside of Japan (due to the specific protocol use it wasnt compatible with other networks, which made it different from the ARPANET); and the encoding of the Chinese characters was a problem. A few years later, in 1984 JUNET (Japan University NETwork) was established, with a connection over a telephone line to the Tokyo Tech University, Keio University and Tokyo University. This network was supported by KDD Lab (now KDDI Lab) and NTT Musashino Lab, both being technology research centers. The first Internet Exchange built in Japan was in 1988, started as a project by WIDE (Widely Integrated Distributed Environments) to experiment with new computers and communication technologies. The first ISP business was started by AT&T Jens and IIJ (Internet Initiative Japan) in 1993 as the first Japanese Commercial Internet Service Provider. In 1997 the first commercial Internet Exchange was started, with the name JPIX with the purpose of providing neutral internet traffic exchange services for ISPs and Content Service Providers so they can exchange traffic economically and effectively. In 1994 NSPIXP was firstly launched as a Japanese IXP, designed and operated by WIDE. Over the years they have led to the current state of development of the commercial IX Service. 4.3.3 India In 1992 the National Centre for Software Development (NCST) assigned [12] and maintained the .IN Top Level Domain (TLD). However after some time these duties were transferred to the Centre for Development of Advanced Computer (C-DAC). This process of getting an .IN TLD was difficult, expensive and a very long process with restrictive eligibility requirements. During this time there were around 6.600 Domain names in the database. In 2004 the government of India decided to transfer the registration of the .IN Domain Name from the National Centre Of Software Technology to NIXI, the very first Internet Exchange in India, which was established in 2003. This was done with help from the Internet Service Providers Association of India (ISPAI) and the Department of Information Technology (DIT). NIXI is a non-profit company. Today NIXI has 30 major ISPs connected, with 67 connections at 7 different locations. At this moment there are about 50.000 Domain Names registered, with 57 Registrars appointed across the world. Qualified people have been assigned to reduce conflicts and maintain the management of the domain names. They have an Internet penetration [5] among their population of 6.9% and make up 9.8% of the total Internet users in Asia. This doesnt sound like a lot, but keep in mind that China already holds 50.9% of the total Internet users of Asia and Japan has only slightly more Internet users than India with 12% of the Asian total.
19
4.3.4 Nepal The Internet market in Nepal [13] is growing rapidly. With less than 1 percent of Nepals population using the Internet in 2005, this number is growing fast. A competitive market between ISPs and low prices to Internet access are the cause of this. Right now 31 private ISPs offer Internet to businesses and home-users, but only two of them dominate the market with a total share of over 70%. These ISPs are WorldLink and Nepal Telecom. An important source to Internet access for the Nepalese are cybercafs, it is believed to have one of the highest concentration of cybercafs in the world. The Kathmandu Valley region has a higher penetration of Internet access, because its more developed while being hard to gain access to the mountain regions where low incomes populations can be found. The Nepali Wireless Networking Project is running an initiative to make Internet access possible in the rural populated mountain regions. In 2002 the first Internet Exchange was established in Nepal, called npIX. Currently they have peak traffic of over 100Mbps with 17 connected ISPs, including the two most dominate ISPs. Nepal has an Internet penetration [5] of 2.2%, and make up about 0.1% of the total users in Asia.
The Internet penetration in Latin America amongst its population is 39.5%. They hold 8.0% of the total Internet users in the world. However in the past 10 years, there has been a gigantic growth, in users, so Internet is becoming more and more common in this region. Of all the Latin American countries, Brazil is by far the largest of the Internet users with some 48.5% of the continents total and the Internet user penetration is at around 37.8%. The second biggest contributor to Internet usage in Latin America is Argentina, with 17% followed by Colombia with 13.7%. However, the Internet penetration amongst the population is significantly higher in Argentina, compared to Brazil, with 64.4% and Columbia with 48.7% respectively. 4.4.1 Brazil In 1962 a research foundation was established in the Sao Paulo region in Brazil, which was given the name FAPESP (Research Foundation Sao Paulo) [14]. In 1988 the ANSP (Academic Network Sao Paulo) was created as an academic network by FAPESP, which would be the first Brazilian network to be integrated with the Internet in 1991 and connected universities and research institutes with one another. The ANSP operated NAP do Brasil, in 2004 Terremark and FAPESP decided to work together by letting Terremark operate the NAP, this would benefit both parties concerned and the general development of Internet in Brazil. Besides FAPESP an Internet Steering Committee (CGI.br) was created in May 1995, with the aim of promoting and improving the participation of the society towards the implementation, management and usage of the Internet. One of the executive arms (Center of Study and Research in Network Technology and Operations, CEPTRO) of the Internet Steering Committee is responsible for projects and services that are aimed to improve the quality of Internet in Brazil. One of their projects was the creation of PTT Metro, which are the Brazilian
20
Internet Exchange Points, located in all of the major Brazilian cities. This project was started in 2004. The most important features of the PTT Metro are to be neutral and independent from commercial suppliers and offer high quality service. Today they have some 14 IXPs with an aggregated peak traffic of more than 30Gbps and more than 200 participants. The Internet Steering Committee is responsible for operating and managing the Brazilian Internet backbone. 4.4.2 Argentina In 1989 an association was formed by: ISPs, Datacenters, Solutions & Content providers and many other entities. This association was named CABASE and created NAP de CABASE in 1997. The purpose of NAP de CABASE was to provide a place where all Internet players Figure 5.4 shows the Latin American could peer efficiently. There were a countries that have one or more IXPs. few things to be considered before this The white countries have no known could be accomplished. The location IXPs operating within them. of the site was very important. It was eventually decided to locate it in Downtown Buenos Aires, the largest and most important city in Argentina and has many Internet players. The decision of the location was also based on the network availability for accessing NAP de CABASE. The next consideration was the model of the NAP. It was decided to use a cooperative and self-governed model. Any cost or investment would be distributed amongst the members of NAP de CABASE. Operating agreements as well as procedures are developed and voted on by the members. Each member has the right of one vote, regardless of its size. NAP de CABASE is also carrier neutral, so even today the site and infrastructure are not owned by any of their member, which minimizes the risk of being affected by problems of a member. The NAP has special agreements for either non-carrier or non-ISP members. They can only connect if their businesses dont affect any of the current ISPs or carriers that are connected. The average number of members connected to NAP de CABASE is 35. Since the beginning of the NAP multilateral peering was the only peering agreements possibility made available, but since two years ago, bilateral agreements have been permitted, but not
21
actively promoted. In November 2010 NAP de CABASE operated the first regional NAP in the Neuqun Province, starting with 7 members. It has a connection to the NAP in Downtown Buenos Aires with access to all the members and traffic there, with the same conditions as their members. This creates a huge cost improvement for the Argentina region. Plans for other regional NAPs have also been made, however no further details have come to hand. After the deployment of the first NAP de CABASE a growth in local hosting, content and e-commerce services was seen, this resulted in more local content becoming available.
4.5 Africa
In
the
last
couple
of
years
the
Internet
usage
in
Africa
has
grown
tremendously,
to
such
an
extent
that
the
current
network
capacity
cant
keep
up
with
the
demand,
which
actually
led
to
falling
behind
on
their
growth.
Right
now
they
could
be
considered
as
being
almost
10
years
behind
on
Europe
and
North
America,
however
this
difference
only
looks
set
to
grow
bigger
and
bigger
over
the
next
couple
of
years.
In
the
last
couple
of
years
the
existing
connectivity
was
mainly
over
satellites.
This
is
very
expensive,
usually
slow
and
sensitive
to
traffic
congestion.
Depending
on
the
part
of
the
African
[15]
region
that
you
are
looking
at,
the
growth,
state
of
Internet
connectivity
and
access
is
very
diverse.
Some
countries
are
well
developed,
while
others
are
very
poorly
developed.
This
makes
it
very
hard
to
connect
all
the
countries
to
each
other
and
provide
better
Internet
connectivity
and
access
within
the
African
regions.
One
of
the
main
reasons
for
the
poor
growth
is
poor
infrastructure
and
very
fragile
end-to-end
connectivity
(for
example,
connectivity
from
the
local
ISP
to
the
home-user),
this
is
increased
by
a
lack
of
adequate
network
backup
routes,
in
case
the
power
goes
down
or
a
cable
is
broken.
Other
disadvantages
for
Africa
are
lack
of
power,
high
import
taxes
(to
get
routers
and
switches
into
the
country),
lack
of
proper
skilled
technical
staff
(they
tend
to
leave
to
more
developed
regions
like
North
America
or
Europe,
once
they
are
skilled),
corruption
and
monopolies.
International
companies
and
organizations
have
started
projects
to
deploy
proper
Internet
connectivity
to
a
larger
number
of
people
throughout
the
region.
Today,
most
of
the
African
traffic
is
routed
through
Europe
or
North
America
even
if
the
destination
of
the
request
is
within
African
borders.
Some
of
the
traffic
even
needs
to
go
through
both
Europe
and
North
America
to
then
go
back
to
the
African
region.
High
international
bandwidth
prices
are
the
biggest
contributor
to
African
ISPs,
a
lot
of
the
traffic
goes
over
satellite
links,
because
fibre
links
are
limited
and
expensive
to
deploy.
High
traffic
costs
and
the
poor
connectivity
can
be
improved
by
building
regional
and
national
IXPs,
this
also
reduces
the
usage
of
international
traffic
lines.
The
few
countries
in
Africa
that
do
have
access
to
international
fibre
dont
feel
the
benefits
of
using
this,
because
they
are
still
charged
at
monopoly
level
prices
by
state
owned
operators,
which
still
dominate
most
of
Africa
and
in
many
other
developing
regions,
so
satellite
links
are
used
instead.
This
is
beginning
to
change
as
many
countries
revise
their
competition
frameworks
as
exclusivities,
granted
to
incumbents,
come
to
an
end.
In
2003
there
were
10
countries
of
the
53
in
the
region
that
had
an
IXP,
in
2006
this
number
had
risen
to
16
countries.
22
Today the total number of IXPs in Africa is 18 and this number looks to grow further. As more and more routes stabilise, more customers will subscribe to ISPs services. In time this will result in better QoS with a wider impact on the deployment of infrastructure in the regions. Local people are demanding better connectivity and the ISPs have no choice but to invest, so as to make more money. This is good for the local economy. Today fibre cables are being deployed right across Africa, however this is a long process where a lot can go wrong, such as sabotage on landlines that are, either broken or stolen, which keeps the incumbents from deploying faster. Some try to solve this by providing wireless access to their users, but this has other implications, which are out of the scope of this research. In the next couple of years the Internet connectivity should have improved from where it is today and it should be more accessible to everybody. At this moment some 14% of the total world population [5] is located in Africa, but only 10% of the total African population has access to Internet. This means there is a lot of potential to grow and create local Internet connectivity to home users compared to that of any other region in the world. Figure 5.5 shows the African countries that have an IXP. The white countries do not have an IXP. Some of the African countries have more than 1 IXP, however this is limited to just two countries, Egypt en South Africa.
23
5 IXP models
There are approximately 300 IXPs [16] around the world, however there are quite a number of differences between these IXPs. No IXP is the same, they may look quite a bit alike, but things will probably be just a bit different in each case. This section will take a look at the differences between the IXP models [16].
24
came from, however sometimes the ownership develops into a more formal body over the years. When this development occurs it usually leads to an IXP becoming Association based. Where a Board has been created who make major decisions about the future of the IXP. It is very common that the members of an association based IXP have voting powers and has some kind of influence on the future of their IXP. Another common body is the business unit. This type of IXPs aims to make a profit and usually provide the collocation space if the decisions made by the IXP management are aimed at making profit which they generally do by selling as much rack space as possible and tend to set their prices based on the market. 5.1.3 Fees There are a lot of differences between fees at IXPs. Some IXPs even claim to be completely free, meaning that they do not charge a port connection fee however they still may require the participants to pay for rack space and local transport to the IXP. In general, this is done by very few IXPs today. Depending on whether an IXP is commercial or non-commercial may also influence the level of the fees. Usually when an IXP is non-commercial you see a monthly fee you have to pay for access to your port, sometimes you also have to pay a connection or establishment fee. This is all really different per IXP. When looking at commercial IXPs, you can see a lot of differences in pricing as well. One of the reasons for this is that most of the commercial IXPs can do their pricing based on the market. This makes them more flexible in a way. Some IXPs want you to pay a one-time fee, and a monthly fee for your port. Some give the port away for free, because you are required to rent rack space in one of their colocation facilities, or dependent on certain services they offer. A very important factor when looking at the pricing between IXPs is whether they have volunteers working for them, or they are properly staffed. When you want to be sure of support, the decision is usually made to choose for dedicated personnel. It is also important to look at the equipment in this case, usually when an IXP is ran by volunteers, the equipment is usually funded by donations from companies, organizations and individuals, while a staffed IXP probably has more budget for equipment. Which means they are less dependant on third parties. 5.1.4 Scope of activities Which services an IXP offers to its participants is really dependent on the scope of their activities. These activities were decided upon when the IXP started, or when a We will see where it goes from there mentality didnt work anymore. What exactly is offered by an IXP is yet again dependant on whether it is commercial or non-commercial. Commercial IXPs usually provide their customers with colocation, hosting, security, customer support, SLAs etc. All these services are also used to make more profit. Even when looking at non-commercial IXPs the scope of activities also tends to differ somewhat, from just offering ports to full customer support, SLAs and rack space.
25
5.1.5 Peering agreements and policies Peering can be achieved on different levels, it can be implemented in a technical way, or it can also be implemented in a non-technical way, based on the peering policies from connected IXP participants. When its based on the peering policies of the IXP participant, it has nothing to do with the IXP itself, because its purely an agreement between two or more parties that connect to the IXP and describes how they want to exchange the traffic and under which conditions this has to be done. There are two ways to implement peering on a technical basis. This can either be done via private peering or through public peering where the traffic goes over an IXP. A definition of both can be found in the table below. Private peering Physical direct connection between only two parties Public peering (over IXP) Physical connection to a medium, managed by a third party, through which multiple parties are connected and approachable Table 6.1: Different ways to implement technical peering. The other possibility of implementing peering is in a non-technical way, where agreements are made between two or more parties. Every connected party to an IXP has a specific peering policy, in which it defines how they want to exchange traffic and under which conditions. The different peering policies are described in the table below. Every connected party can have one of these policies, based on their business model and peering conditions. This peering policy may differ per region. For example the local incumbent in a country in Europe can choose to have a restrictive peering policy in Europe, however when this incumbent expands it business to a different region, for example North America, it may decide to have an open peering policy as this may make more financial sense. The connected parties at IXPs in North America might be more interesting to peer with than the connected parties at the IXP in its home country, where they feel it will decrease their business. Policy Definition Open Peering Generally willing to peer with anyone, with no or few prerequisites. Selective Peering Generally willing to peer, however there are some prerequisites, once met this generally leads to peering. Restrictive Peering Generally not willing to peer extremely difficult to meet prerequisites with high possibility of denial once they are met. No Peering No intention to peer. Traffic is exchanged via transit agreements Table 6.2: Peering policies for IXP participants. The implementation of the peering policies can be done in a number of different ways. This happens on a non-technical way and is more agreement based on the
26
conditions made between different parties. The connected parties can implement exchanging traffic through buying transit, having a bi-lateral or multi-lateral peering agreement. The distinction between the three can be found below. Transit Gives access to the complete Internet Bi-Lateral Gives access to the networks of the parties you are peering with and their client-networks. Multi-Lateral Access to multiple networks from parties you are peering with, based on one agreement Table 6.3: Possible ways to implement peering policies. A current popular trend is paid-peering, where one party pays another party to peer with them, this is usually done on the difference in exchanged traffic between the two parties. This is a modification of the bi-lateral peering agreement. The term Mandatory Multi-Lateral peering is often used when every connected party at an IXP is forced to peer with every other party, this peering agreement can be seen as multi-lateral peering, where all connected parties agree to exchange traffic under the same agreement. If they do not agree to this agreement, there will be no peering available for them. Figure 6.1 shows a transit agreement. Network A buys transit from its Transit Provider, which provides Network A with connectivity to the Internet. .
27
Figure 6.2 shows a Bi- lateral peering agreement. Network C buys transit from Network B, when Network A and Network B have agreed to peer (bi- lateral peering), Network A will get access to Network C. Network C will also get access to Network A, because of the transit agreement between Network B and Network C.
Figure 6.3 shows a Multi- lateral peering agreement. Network B, D, E and G have agreed to peer under the same peering conditions, so they have access to each others networks. Network C still buys transit from Network B Network A and Network B still have a bi-lateral peering agreement, this provides network A with access to Network D, E and G and the other way around.
28
5.2 Europe
When
looking
at
the
different
IXPs
in
Europe,
you
can
see
that
quite
a
large
proportion
of
IXPs
are
of
a
non-profit
nature
and
most
tend
to
be
both
colocation
and
carrier
neutral.
The
main
reasoning
behind
this
neutrality
is
that
the
ISPs
are
more
willing
to
join
an
IXP
that
isnt
controlled
or
influenced
by
a
third
party.
They
maintain
this
neutrality
by
being
located
at
one
or
more
datacenters
that
are
owned
by
separate
entities.
Thus
no
special
treatment
is
available
for
any
one
participant
nor
the
datacenter
companies.
Besides
the
neutrality
there
are
also
some
differences
in
ownership
of
the
IXP,
however
most
IXPs
in
Europe
are
association
based
or
are
run
by
an
academic
institution.
There
are
a
small
number
of
commercial
IXPs
in
Europe,
however
this
is
generally
limited
to
the
smaller
to
mid-sized
IXPs.
There
is
a
trend
in
some
countries
such
as
The
Netherlands,
Italy
and
France
for
region
al
governments
to
establish
local
IXPs,
however
to
date,
these
havent
grown
to
be
all
that
significant.
In
general
most
of
the
IXPs
in
Europe
just
stick
to
their
main
goal
of
renting
ports
on
their
switching
infrastructure,
however
there
is
a
trend
of
late
that
has
seen
some
not-for-profit
IXPs
extending
their
services
to
fulfil
member
needs.
The majority of IXPs in North America are of a commercial nature. They generally dont use the word IXP nor NAP in their corporate name, which already suggests that being an IXP/NAP is not their primary goal. Their main aim goal is to make a profit by selling rack space to their customers. They also sell additional services like security, hosting and SLAs as an extra source of income. Being a datacenter makes interconnecting for two of the IXPs customers relatively easy, because only a physical fibre needs to be laid between the two AS networks. Quite a number of IXP customers locate themselves within a datacenter, to make interconnecting with other customers easier and to have better connectivity to other Autonomous Systems as the distance to other customers is relatively small at a datacenter. The huge competition between IXPs in North America makes pricing and selling the ports interesting. For example, they make it enticing for their customers to buy a 10Gbps port, compared to a 1Gbps port, by selling them very cheap or giving them away for free. However this leads to the customer having to buy more rack space as well and probably other extra services, before they can operate their port. Some non-profit IXPs can also be found in North America, a good example is SIX (Seattle IX), which is a neutral IXP with volunteers as their staff. Connected parties are dependant on the time the volunteers are willing to spend. The equipment used by them is based purely on donations. They are completely dependant on the donations of business, organisations and the time of their volunteers.
29
5.4 Asia
The
Asian
Pacific
region
is
usually
thought
of
as
a
closed
community
where
interconnecting
is
hard
to
maintain
and
setup.
This
may
be
true,
however
its
slowly
becoming
easier
to
interconnect
in
Asia.
One
of
the
biggest
problems
the
Asian
market
is
struggling
with
is
the
artificially
high
price
of
transit,
transatlantic
costs
and
interconnecting
costs
at
IXPs.
Most
of
the
dominant
ISPs
still
think:
for
you
to
get
more
routes
to
the
Internet
you
have
to
buy
transit
from
me.
This
is
only
partly
true.
It
works
for
the
ISPs,
however
it
doesnt
help
achieve
getting
better
connectivity
in
the
country
nor
the
region.
The
idea
of
peering
does
not
interest
them,
either
due
to
the
fact
that
they
dont
know
the
benefits
of
connecting
at
an
IXP
or
are
afraid
of
losing
business.
This
keeps
the
prices
of
Internet
traffic
very
high
for
smaller
ISPs
and
overseas
companies.
Creating
an
IXP
would
be
easy,
however
a
lot
of
parties
dont
see
the
benefit
from
doing
this,
because
of
the
tremendous
high
prices
of
connecting
to
an
IXP.
Asia
is
also
troubled
with
a
lot
of
regulated
telecoms.
Prices
in
de-regulated
regions
are
generally
lower
than
regulated
regions.
Many
countries
in
Asia
are
dependant
on
neighbouring
countries
for
transit
connectivity.
Another
problem
in
Asia
are
culture
differences,
To
help
solve
all
of
the
above
mentioned
problems
its
important
for
all
involved
parties
to
participate
in
such
forums
as
APNIC,
SANOG
and
Apricot.
This
helps
to
create
a
better
understanding
of
the
concept
of
IXPs
and
peering,
but
also
gives
people
the
opportunity
to
exchange
ideas
and
talk
about
problems
with
more
experienced
people
from
all
over
the
world.
Its
also
a
good
source
for
future
customers
and
peering
relations.
Both
commercial
as
non-profit
IXPs
can
be
found
throughout
Asia.
However,
most
of
the
more
commercial
IXPs
can
be
found
in
wealthier
regions,
like
China,
Japan
and
Australia,
while
less
developed
regions
like
India,
Nepal
and
the
Philippines
tend
to
have
more
non-profit
IXPs.
30
traffic goes via Miami and back because its cheaper than deploying fibers across the multiple countries.
5.6 Africa
The
African
region
is
poorly
developed
compared
to
other
regions
around
the
world.
This
makes
it
very
interesting
to
take
a
look
at
how
they
are
developing
and
building
a
peering
ecosystem.
All
IXPs
that
are
currently
operating
in
Africa
are
non-profit
neutral
IXPs,
this
is
probably
also
the
only
possible
way
to
get
Internet
connectivity
up
and
running.
The
threshold
for
ISPs
is
significantly
lower
to
connect
with
each
other
when
its
a
neutral
exchange.
The
main
reason
would
be
the
already
existing
(non)
trust
issues
that
are
going
on,
doesnt
make
it
much
easier
to
start
business
with
each
other.
Most
of
the
IXPs
in
Africa
are
established
either
by
the
Government
or
a
local
organization
of
Telecommunications.
There
are
a
few
countries
in
Africa
that
are
well
developed
compared
to
other
countries
in
the
region.
Nigeria
is
one
of
the
best-developed
countries
in
Africa
as
far
as
Internet
penetration
goes.
Egypt
has
the
2nd
best-developed
Internet
penetration
in
Africa.
For
the
regions
in
Africa
that
dont
have
an
IXP,
special
initiatives
have
been
set
up
to
establish
IXPs
with
the
local
ISPs
and
provide
connectivity
between
African
countries.
These
initiatives
are
supported
by
international
companies
and
organizations,
which
work
together.
Over
the
next
few
years
a
huge
increase
of
African
traffic
can
be
expected
due
to
the
establishment
of
regional
IXPs
and
the
deployment
of
further
Internet
connectivity.
31
Currently a lot of developments are taking place in the world of the Internet. These changes, trends and developments are of importance to the future growth of IXPs around the world. This section will take a closer look at what developments and trends are occurring.
32
over Private Interconnections, it is also likely that this traffic will not return to the public fabric of the IXP. They also mention that IXPs should make sure they keep their current level of quality or even improve on this. The non-profit IXPs should work on transparency and IXPs should try to work together with Euro-IX to standardize service offerings to their participants, like IPv6, multicast, (virtual) PIs, etc.
6.2 Trends
The 2010 Euro-IX Report [17] on European IXPs shows some trends that are occurring in the IXP world, especially in the European region. One of the biggest changes since to last year is the huge increase of peak traffic at IXPs in Eastern Europe. Countries like Russia, Bulgaria, Ukraine, Romania and Latvia have grown tremendously and rapid developments can be expected over the next couple of years with a real contribution to the total European IXP traffic aggregate. Other regions, which are also developing rapidly, include Latin America (especially Brazil and Argentina) and the African region however lots of investments are needed to improve the infrastructure, although a huge potential is present. A large growth can also be seen in the number of unique ASNs that are peering at two or more IXPs in Europe. Between 2008 and 2009 this number had grown by about 20%, while from 2009 to 2010 this has grown some 24%. Between 2007 and 2008 certain growth factors of IXPs in Europe slowed down or even decreased somewhat. For instance the amount of traffic per unique ASN only slightly increased, as did the number of known operating IXPs, while the average traffic load per switch increase dramatically. While 2008 2010 saw strong growth in nearly all aspects of the European IXP scene. It is expected that the Eastern European region will maintain a strong growth rate in the coming years. When comparing the categorized and regional IXP annual growth rates, from 2009-2010, one can see the largest growth both in Medium and Small IXPs as well as those IXPs coming from the Southern Europe and Northern Europe regions. Comparing the global growth rated from 2009 and 2010 one can see that both European and North American IXPs have grown on average between 61% and 67%, Asia on the other hand saw growth rates of around 15% compared to those of last year, while Brazil saw growth rates of around 138% and the Global average was at around 57%. Another trend worth noting is the growth in route server usage by IXPs in Europe. At this moment 19 IXPs are known to be operating a route server, however it is estimated that more IXPs are in fact making use of them and it is expected that more IXPs will follow suit in the coming years. It is also apparent that the core business of the participants, connected to an IXP, has changed over the past 5-10 years. While a company used to cover just one particular area of business (for example a pure ISP), that same entity now operates in 2, 3 or more different areas of business (e.g. ISP, hosting provider, content provider etc). It is further worth noting that transit prices are still falling dramatically and it is uncertain what influence this will have on IXPs in the future. It might be possible that more traffic will go over transit lines, however it is also possible that the growth of traffic going over IXPs will continue. If over the next couple of years
33
the traffic exchanged at IXPs keeps growing at the current rate, this will result in IXPs having to upgrade their platforms to keep up with this fast development. It is also anticipated that IXPs may also offer more and a broader range of services to their connected participants, however this will naturally be different from IXP to IXP. 6.2.1 Japan A few years ago it was stated [18] that Japanese traffic stays for about 80% within Japan and about 5% of traffic destined to/from the US with estimated traffic of between three and four Gigabits per second. This would make Japan one of the more self-contained self-sustaining peering ecosystems. After calculations made from the statistics of the Ministry of Internal Affairs and Communications of Japan [19], it appears that this situation was indeed correct around 2005-2006, but the Internet situation has changed since then. May May May May May May 2005 2006 2007 2008 2009 2010 Inside 85% 78% 76% 75% 74% 69% Japan To 15% 22% 24% 25% 26% 31% outside Japan Table 7.1: Upload traffic from customers. May May May May May May 2005 2006 2007 2008 2009 2010 Inside 80% 74% 68% 67% 66% 57% Japan To 20% 26% 32% 33% 34% 43% outside Japan Table 7.2: Download traffic from customers. Above calculations represent only a fragments of the Japanese traffic from 6 ISPs.
34
7 Differences
As far as IXPs go, there are some major differences between the many regions of the world, however these differences can also extend to within a region. This section will cover some of the differences that are of importance, and some further questions have been raised by Euro-IX.
Note: Data for the empty fields was not available in that specific year. As can be seen from the table, a very large amount of the European IXPs have Content Providers and Dedicated Hosting and Governments are also very often connected to an IXP. Its important to note that in a country there are often multiple IXPs, parties such as Governments or Enterprises might decide not to connect to every single IXP in their country. As previously mentioned over the last 5-10 years a trend can be seen in IXP participants. Most participants have expanded their business, for example, a search engine 10 years ago, was just a search engine, however this same entity now also provides hosting or is a content provider and provides an email service
35
etc. This goes for a lot of participants; this also makes it harder to categorize the participants in any manner.
36
IXP Participants 5829 IXP Participants 1116 IXP Participants 2055 IXP Participants 313 IXP Participants 202 IXP Participants 9515
Europe Unique ASNs Advertised ASNs 3181 15685 Asia-Pacific Unique ASNs Advertised ASNs 647 4354 North America Unique ASNs Advertised ASNs 849 13979 South America Unique ASNs Advertised ASNs 186 1538 Africa Unique ASNs Advertised ASNs 114 436 Global Unique ASNs Advertised ASNs 4699 35992
% of Adv. ASNs 20.28 % of Adv. ASNs 14.86 % of Adv. ASNs 6.07 % of Adv. ASNs 12.09 % of Adv. ASNs 26.15 % of Adv. ASNs 13.06
Note: Numbers were taken on 20-12-2010 [16] In this table the IXP Participants are the total number of participants at the IXPs per region, when a participant is connected to multiple IXPs, it will be counted multiple times, since for every IXP this is a different participant. The Unique ASNs are the total number of unique ASNs at all IXPs in a particular region, so an ASN present at multiple IXPs in a region, is only counted here once. When comparing North America and Europe to each other, one can see that there are quite a number of differences present. For instance, in the U.S. one of the main ones would be the fact that the U.S. consists of states, however they belong to one country. While every state might have slightly different laws, they all speak the same language, generally require the same procedures to get a company started and often the people residing in the states have very similar needs to one another. This makes it easier for a company to expand to different states of the U.S. and run their services from there, because there will be less fuss with new laws and taxes etc. In Europe however this is completely different. Every single country has its own set of laws, language, culture, specific rules,
37
cultural needs and more. All of this makes it a lot harder for a company to expand to a new country. As can be seen in Figure 7.2 (Map of the U.S. states with an IXP), there are a large number of states that have no IXPs at all. This means that the physical location between cities (in the states) that already have an IXP (the ones without are probably less interesting for a company to establish themselves and do business) can be quite large, especially compared to Europe, where most of the larger cities have at least one IXP. This makes it more attractive for companies or organizations to expand to different countries. The costs of fibre will probably be less expensive, because the distance between two cities is relatively close, compared with those in the U.S. This makes it more expensive for companies to reach out to an IXP, let alone multiple IXPs. Especially for smaller companies its cheaper to buy transit instead of connecting to an IXP. The added value of the IXP wont bring enough reward to their business. When two companies do decide to exchange traffic in the same datacenter, this is usually done via cross connects, because the port prices are more expensive than the physical cable laid between their two networks (this cable is usually just within the datacenter itself, so not so expensive compared to renting a port at the IXP). Large quantities of traffic going between two networks often makes a cross connects more often and more financially interesting, but offering the IXP service can act as a great draw card.
38
8 Conclusion
During the process of researching these differences between IXPs around the globe following was discovered. Commercial IXPs have a dominant position in the US, while in Europe mainly non-profit IXPs can be found, however there are also some commercial IXPs present. In Asia there is a mix between commercial and non-profit IXPs. In more economically developed Asian counties (China, Japan, Australia etc.) it is more common to see commercial IXPs. In the economically less developed countries (Nepal, India, Malaysia, etc.) more non- profit IXPs are present. The main difference between commercial and non-profit is that when an IXP is commercial the colocation owner is also the owner of the IXP. In the US it is more common to use Private Interconnections instead of peering publicly over the exchange (which is more common at non-profit IXPs). In Europe IXPs were created as a way to decrease the costs that otherwise had to be paid to the incumbent to exchange the traffic, while in the US IXPs evolved simply as means of making profit. From what has been researched a few important predictions can be made about the future. Over the next 5-10 years, it is likely that we will see an increase of the number of commercial IXPs in Europe. It is likely that non-profit IXPs will establish in the countries across the world that currently have no IXP at all. Once these non-profit IXPs have been established the country will probably increase to be more interesting for commercial IXPs to establish as well. European IXPs will start to offer more services towards their participants, which overtime might also be the case in (now) less developed regions. It can be expected that more companies and organizations will establish peering relationships with Asian companies, though this might develop slow due to the slowly dropping prices of transpacific transit and colocation in Asia. Finally, there is no doubt that we will see a huge increase in Internet bandwidth, which will require platforms at IXPs to be upgraded to keep up with the demand.
39
Glossary of terms
Autonomous System (AS): a (large) network on the Internet that is used to distribute traffic to and from other networks. Autonomous System Number (ASN): number that is used to identify an Autonomous System. Border Gateway Protocol (BGP): Internet protocol that is used to announce tables, which hold IP address ranges that can be reached through a specific AS. Colocation: facility from where an IXP or datacenter can operate. They can have multiple of these facilities that are all connected to the same network. Often known as a datacenter facility. Content Provider: company or organization that makes content like video, social media, music accessible to the Internet. Incumbent: Telecom Company who used to have monopoly in a specific region or country. Interconnection: physical connection between an AS and equipment or a facility that does not belong to them (for example an IXP or datacenter). Internet Access Provider (IAP): company that offers Internet access to its customers. Often referred to as an ISP. Internet Exchange Point (IXP): Physical point of infrastructure where parties can connect to and exchange Internet traffic with each other. Internet Service Provider (ISP): Company that offers Internet access to its customers using techniques such as DSL, Cable, or FTTH. Currently most ISP's offer services like e-mail access, VoIP, Television and other additional services. Private Network Interconnection (PNI): A physical cable between two Autonomous Systems to facilitate private peering. Route server: a centralized server provided by an IXP that collects tables with IP address ranges and distributes them to other peers.
40
Acknowledgements
Special thanks to everybody who took time into answering my, often many (annoying) questions, requests and e-mails in general. You have all been very useful! I could write down a list of all people who have helped me out here, with the risk of forgetting names or making people upset if they do not wish to be named. People from the following companies, organizations and Forums have contributed to this document: - 17th Euro-IX Forum attendees - NANOG50 attendees who were willing to share their business cards - Diverse number of people participating in the Asian IXP scene - Akamai - Amazon - APNIC - Equinix - Hurricane Electric - IPLAN - JPIX - JPNAP - LACNIC - Limelight - PhoenixNAP - Ripe NCC - Telx - Yahoo Of course also all the people that I forgot to mention above. All your help has been very much appreciated!
41
Bibliography
[1]
[2]
[3]
[4]
[5]
[6]
[7]
[8]
[9]
Random
House,
Dictionary.com,
definition
of
Internet,
the
(accessed
01-11-2010)
http://dictionary.reference.com/browse/internet
Y.
Rekhter
et
all.
A
Border
Gateway
Protocol
4
(BGP-4),
2006
http://www.ietf.org/rfc/rfc4271
J.
Hawkinson
et
all.
Guidelines
for
creation,
selection
and
registration
of
an
Autonomous
System
(AS)
1996
http://tools.ietf.org/search/rfc1930
K.
Lindqvist,
Netnod,
IXPs
in
Europe
and
the
Business
case
for
peering,
2009
http://www.menog.net/menog-meetings/menog4/presentations/
ixps-europe-business-case-peering.pdf
Miniwatts
Marketing
Group,
Internet
World
Stats,
2011
http://www.inter
networldstats.com/stats.htm
Amsterdam
Internet
Exchange,
Traffic
Statistics,
2011
http://www.ams- ix.net/statistics/
London
Internet
Exchange,
LINX
Traffic
Stats,
2011
https://www.linx.net
/pubtools/trafficstats.html
Deutsche
Commercial
Internet
Exchange,
DE-CIX,
2011
http://de- cix.net/content/network.html
US
National
Science
Foundation,
The
launch
of
NSFNET,
http://www.nsf.
gov/about/history/nsf0050/internet/launch.htm
[10] C. Cheng, Internet Exchange for local traffic: Hong Kongs Experience, http://www.isoc.org/inet96/proceedings/h1/h1_3.htm [11] H. Esaki, WIDE Project Activities, 2000 http://hiroshi1.hongo.wide.ad.jp/ hiroshi/wide-wp/China/wide.html [12] A. Singhal et all. National Internet Exchange of India (NIXI), 2009 http:// www.apricot.net/apricot2009/images/lecture_files/nixi_apnic_amitabh_s inghal_feb09.pdf [13] OpenNet Initiative, Nepal, 10-05-2007 opennet.net/sites/opennet.net/file s/nepal.pdf [14] Pesquisa FAPESP Online, Strategic Partnerships, 03-2002 http://revistap esquisa.fapesp.br/?art=509&bd=1&pg=1&lg=en
42
[15]
W. Stucke, Challenges facing ISPs in Africa: a view from an African ISP. 2008, http://www.afrispa.org/dpages/Powerpoint/Challenges%20facings %20ISPs%20in%20Africa.ppt
[16] European Internet Exchange Association, Euro-IX, http://www.euro- ix.net/ (see note below) [17] S. Radovcic, European Internet Exchange Association 2010 Report on European IXPs, 2010 http://www.euro-ix.net/resources/reports/euro- ix_report_2010.pdf [18] B. Norton, Japan Internet Peering Ecosystem, 2005 http://peering.drpeeri ng.net/white-papers/Ecosystems/Japan-Peering-Ecosystem.php [19] Ministry of Internal Affairs and Communications, Understanding of the total Internet traffic in our country, 2010 http://www.soumu.go.jp/main_ content/000080807.pdf (Japanese) [20] G. Huston, Distribution Reports, AS Resource Allocations, 2010 http://resources.potaroo.net/iso3166/archive/ Note: Some statistics from Euro-IX are only accessible with login credentials.
43
Appendix
Since
2006
statistics
on
all
IXPs
in
every
European
country
were
gathered
by
Euro-IX.
Together
with
other
resources
[20]
a
table
has
been
put
together
to
be
able
to
show
the
growth
in
ASNs
at
IXPs
from
2007
to
2010.
Due
to
missing
statistics
on
ASNs
in
2006,
no
growth
can
be
seen
from
before
2007.
The
column
Unique
ASNs
indicates
the
total
unique
ASNs
that
are
present
at
European
IXPs.
The
column
Allocated
ASNs
indicates
the
total
number
of
allocated
ASNs
in
this
country,
Advertised
ASNs
shows
the
total
number
of
advertised
ASNs
in
a
country
and
the
column
%ASNs
at
IXPs
shows
the
total
percentage
of
ASNs
that
are
present
at
IXPs
in
a
country.
2007
Country Austria Belgium Croatia Cyprus Czech Republic Denmark Estonia Finland France Germany Greece Hungary Iceland Ireland Italy Latvia Luxembourg Malta Netherlands Norway Poland Portugal Romania Russia Slovakia Slovenia Spain Sweden Switzerland Ukraine United Kingdom Total Unique ASNs 90 79 17 0 72 0 17 26 220 318 15 47 14 33 89 118 13 0 353 63 64 19 41 206 44 0 52 82 111 81 415 2699 Allocated ASNs 309 116 52 47 137 142 23 121 486 1187 141 175 22 81 498 150 18 24 396 106 755 59 911 1867 61 110 255 351 363 1046 1446 11455 Advertised ASNs 245 89 41 28 115 97 20 89 329 859 93 119 17 55 374 114 14 20 290 70 631 40 657 1478 48 91 190 232 274 821 969 8509 % ASNs at IXPs 36.73% 88.76% 41.46% 0.00% 62.61% 0.00% 85.00% 29.21% 66.87% 37.02% 16.13% 39.50% 82.35% 60.00% 23.80% 103.51% 92.86% 0.00% 121.72% 90.00% 10.14% 47.50% 6.24% 13.94% 91.67% 0.00% 27.37% 35.34% 40.51% 9.87% 42.83% 45.26%
2008
Country Austria Belgium Croatia Cyprus Czech Republic Denmark Estonia Finland France Germany Greece Hungary Iceland Ireland Italy Latvia Luxembourg Malta Netherlands Norway Poland Portugal Romania Russia Slovakia Slovenia Spain Sweden Switzerland Ukraine United Kingdom Total Unique ASNs 105 78 17 7 85 0 16 29 233 388 15 50 15 36 99 118 13 0 394 69 69 20 62 232 48 15 54 90 136 127 458 3078 Allocated ASNs 318 125 58 48 161 166 29 122 526 1233 151 200 28 87 538 174 20 24 427 124 872 63 956 2279 66 136 286 384 390 1271 1613 12875 Advertised ASNs 255 103 46 31 135 120 22 96 372 898 100 139 23 67 414 134 16 19 314 80 742 47 736 1851 54 119 220 267 299 1031 1109 9859 % ASNs at IXPs 41.18% 75.73% 36.96% 22.58% 62.96% 0.00% 72.73% 30.21% 62.63% 43.21% 15.00% 35.97% 65.22% 53.73% 23.91% 88.06% 81.25% 0.00% 125.48% 86.25% 9.30% 42.55% 8.42% 12.53% 88.89% 12.61% 24.55% 33.71% 45.48% 12.32% 41.30% 43.70%
2009
Country Austria Belgium Croatia Cyprus Czech Republic Denmark Estonia Finland France Germany Greece Hungary Iceland Ireland Italy Latvia Luxembourg Malta Netherlands Norway Poland Portugal Romania Russia Slovakia Slovenia Spain Sweden Switzerland Ukraine United Kingdom Total Unique ASNs 111 62 15 6 93 50 18 28 234 470 15 47 15 47 135 158 16 0 447 72 110 20 69 400 47 16 60 102 159 161 493 3676 Allocated ASNs 335 146 75 54 396 187 28 143 602 1259 145 196 29 105 578 207 29 23 458 151 1035 64 1192 2918 79 176 324 421 450 1551 1760 15116 Advertised ASNs 274 122 64 39 275 134 23 114 442 963 105 148 27 83 464 157 24 17 340 101 884 54 883 2435 63 150 253 300 345 1173 1263 11719 % ASNs at IXPs 40.51% 50.82% 23.44% 15.38% 33.82% 37.31% 78.26% 24.56% 52.94% 48.81% 14.29% 31.76% 55.56% 56.63% 29.09% 100.64% 66.67% 0.00% 131.47% 71.29% 12.44% 37.04% 7.81% 16.43% 74.60% 10.67% 23.72% 34.00% 46.09% 13.73% 39.03% 41.25%
2010
Country Austria Belgium Croatia Cyprus Czech Republic Denmark Estonia Finland France Germany Greece Hungary Iceland Ireland Italy Latvia Luxembourg Malta Netherlands Norway Poland Portugal Romania Russia Slovakia Slovenia Spain Sweden Switzerland Ukraine United Kingdom Total Unique ASNs 109 66 17 7 100 50 18 30 262 514 12 52 15 57 165 168 24 12 540 74 155 20 87 465 53 16 66 116 167 174 518 4129 Allocated ASNs 365 157 78 61 666 211 36 158 673 1333 150 198 35 114 617 216 35 25 502 162 1234 64 925 3293 84 204 359 456 543 1720 1862 16536 Advertised ASNs 291 127 66 39 465 154 31 124 486 999 95 158 31 83 481 163 28 15 382 111 983 55 562 2713 72 166 276 324 383 1296 1295 12454 % ASNs at IXPs 37.46% 51.97% 25.76% 17.95% 21.51% 32.47% 58.06% 24.19% 53.91% 51.45% 12.63% 32.91% 48.39% 68.67% 34.30% 103.07% 85.71% 80.00% 141.36% 66.67% 15.77% 36.36% 15.48% 17.14% 73.61% 9.64% 23.91% 35.80% 43.60% 13.43% 40.00% 44.30%
Growth
2007-2010
Country Austria Belgium Croatia Cyprus Czech Republic Denmark Estonia Finland France Germany Greece Hungary Iceland Ireland Italy Latvia Luxembourg Malta Netherlands Norway Poland Portugal Romania Russia Slovakia Slovenia Spain Sweden Switzerland Ukraine United Kingdom Total Unique ASNs 21.11% -16.46% 0.00% 0.00% 38.89% 0.00% 5.88% 15.38% 19.09% 61.64% -20.00% 10.64% 7.14% 72.73% 85.39% 42.37% 84.62% 0.00% 52.97% 17.46% 142.19% 5.26% 112.20% 125.73% 20.45% 6.67% 26.92% 41.46% 50.45% 114.81% 24.82% 37.74% Allocated ASNs 18.12% 35.34% 50.00% 29.79% 386.13% 48.59% 56.52% 30.58% 38.48% 12.30% 6.38% 13.14% 59.09% 40.74% 23.90% 44.00% 94.44% 4.17% 26.77% 52.83% 63.44% 8.47% 1.54% 76.38% 37.70% 85.45% 40.78% 29.91% 49.59% 64.44% 28.77% 50.25% Advertised ASNs 18.78% 42.70% 60.98% 39.29% 304.35% 58.76% 55.00% 39.33% 47.72% 16.30% 2.15% 32.77% 82.35% 50.91% 28.61% 42.98% 100.00% -25.00% 31.72% 58.57% 55.78% 37.50% -14.46% 83.56% 50.00% 82.42% 45.26% 39.66% 39.78% 57.86% 33.64% 51.59% % ASNs at IXPs 1.97% -41.45% -37.88% -20.51% -65.65% -12.99% -31.69% -17.18% -19.38% 38.98% -21.68% -16.67% -41.24% 14.46% 44.15% -0.43% -7.69% 0.00% 16.13% -25.93% 55.46% -23.44% 148.06% 22.97% -19.70% -23.53% -12.63% 1.29% 7.63% 36.08% -6.60% -1.91%
Note: Statistics from the countries indicated in green were not present for certain years, this counts for: Cyprus, Denmark, Malta and Slovenia. The calculations have been made from the first year data was available for that specific country.