This article was originally written as a paper following the 2014 CPRSouth Conference. However, work and procrastination set in, and after half a year of just sitting on it (I completed this in December 2015) I decided to upload it here in the format I know best – a blog. Scroll down past the introduction and the method if you want to skip the process of how I arrived at these names.
It’s generally agreed that the Internet really began in 1982. Of course, there were a few computer networks in the US, the UK and France – ARPANET, CLYCADES, X,25 and the like – but these were very much like the primordial soup from which life supposedly arose: nobody really knows when that bacteria really popped into being. It was when a set of standardized protocols – called TCP/IP – was introduced to the ARPANET and the concept of a worldwide network based on these protocols became an acceptable proposition.
This network, once implemented, grew out of the bones of ARPANET, at first connecting universities to computing resources, and later spreading rapidly to Europe, Australia and Asia. It is paramount to note that this particular network originated in the United States of America.
Today, based mostly on the same TCP/IP protocol suite, the Internet has become a massive influence on human communication and knowledge sharing. BY 2007, over 97% of voice and data traffic was being carried out through the Internet. By 2014, over 2 billion people were connected to the Internet in one way or the other – and many more are expected to gain access everyday.
Note: while the term “the World Wide Web” is often used to describe the Internet, the WWW is simply one the many services to have risen out of the Internet.
However, the Internet – though commonly referred to as a single, abstract body – is very much a collection of private networks connected to each other.
It has since moved far out of government control. It is now supported by a massive collection of privately owned infrastructure that connects equally massive (and also, privately owned) datacenters to each other and to the general public.
Out of these, some infrastructure is more critical than others. This is where the concept of Tier 1 networks and carriers come in.
There is no authoritative definition for a Tier 1 network: rather, it is a term that has evolved over time to classify certain networks that carry the majority of the traffic of the Internet.
Tier 1 carriers typically operate large, high-capacity networks that form “principal data routes between large, strategically interconnected networks and core routers on the Internet” – in essence, backbone networks. They usually have with direct physical links to similar networks and private peering agreements that enable data to be routed through these networks.
Tier 1 ISPs often do not cater to end users, but instead sell internet transit – essentially, the right to allow network traffic through their lines. Providers in the Tier 2 category both peer and purchase transit from multiple Tier 1 networks for their connectivity.
One can surmise that a tier 1 provider is one that have access to the entire Internet region solely via its physically established lines or peering agreements without having to pay a third-party provider for said connectivity.
I can further outline this definition by stating that Tier 2 networks are those that require the purchase of Internet transit to attain global reach. This forms a pyramid with the Tier 1 companies forming the root of Internet connectivity in the world.
These networks are important – not just because they form the backbone of the Internet, but also because they have a very large say in where data goes. As detailed in a paper called “A Fragile Internet: Non-Technical Issues Leading to Internet Blackouts”, the peering agreements that make it possible for global connectivity to exist rely heavily on corporate politics between these giants – there have been cases of companies de-peering, where which one party disconnects the link in to gain the upper hand in a bargain.
An example is the Sprint-Nextel vs Cogent dispute in 2008, in which Sprint-Nextel severed communications between its networks and that of Cogent. This made it impossible for the customers of either network to connect to the other. In fact, Cogent has a history of such disputes – with TelioSonera in 2008, Level 3 in 2005.)
Large companies like these also have the power to change routing paths, impose levies and such, as seen in February 2010, when the CEO of Telefónica (a commonly acknowledged Tier 1 provider for Europe and South America) announced that they intend to charge search engines such as Google for use of their network.
In essence, these companies own this infrastructure own the internet as we know it.
The whole purpose of this blogpost is to list these entities. To avoid scope creep, we will be looking only at commercial infrastructure that is used daily by a massive population.
The closest thing to a complete list of these carriers is on the peering knowledge portal DrPeering.net, a site run by William B Norton, former co-founder and CTO of the Internet Peering Exchange Equinix. Norton has an extensive history with the Internet and extensive research to his name, and his FAQ for “Who are the Tier 1 ISPs” seems to be the most complete list of Tier 1 ISPs available on the web.
However, the list in question appears to list all possible Tier 1 ISPs with no distinction between regional ISPs and truly global networks. Most of the listed ISPs are Tier 2 networks – networks that peer with some networks but also buy transit to reach some portions of the Internet.
A more viable approach is to use Autonomous Systems to generate a more accurate list. An Autonomous System is a network, or a collection of networks that is managed by a single entity. This could be either a single network of a group of networks connecting through one clearly defines routing policy. Each such network is assigned a unique AS number and interconnections between the numbers show the flow of traffic on the Internet.
Autonomous Systems are the key to studying the topology of the Internet; the most interconnected Systems are ISPs that handle large amounts of traffic, and these are clearly highlighted when cast into a visual graph.
CAIDA’s AS Rank, which maps Autonomous Systems to ISPs and ranks them by their customer cone size – the number of customers that are directly and indirectly covered by each Automated System. As CAIDA points out, “ASes with large customer cones play an important role in the Internet’s capital and governance structure. At the top of this hierarchy are ISPs commonly known as Tier-1 ISPs, which do not pay for transit to upstream providers at all; instead they peer with each other to provide connectivity to all destinations in the Internet. At the bottom of the hierarchy are customer ASes who do not have their own customers and pay providers to reach all destinations in the Internet.”
By cross-referencing the CAIDA data with the published network coverage maps of each of the providers, it’s possible to identify global Tier 1 ISPs.
As of the time of writing (December, 2015), the companies listed below appear to be the world’s Tier 1 ISPs, ranked by the number of Autonomous Systems in their customer cone.
Most providers operate one large AS that utterly outranks the rest and gives the provider its CAIDA rank, and therefore for reference, these AS numbers have been attached to the provider’s name. The data is publicly available and can be accessed by anyone to verify.
Level 3 (AS3356, AS3549)
|Number of ASes:||16|
|Customer Cone size||30,814|
Level 3 is the largest network in the world in terms of the number and size of connected Autonomous Systems. Its customer cone covers 79% of all IPv4 addresses and 67% of all ASes – meaning that the Level 3 network can directly connect to and accept connections from just under 80% of the conventional Internet.
A Colorado-based company, Level 3 has been involved in a dispute regarding Comcast and Netflix and have been accused of wiretapping Google, Yahoo and a German Internet Exchange Point on behalf of the NSA. They acquired Global Crossing in 2011, but as of 2014 the networks had not yet been merged. While AS 3356 is the largest here, It should be noted that Level 3 has a second AS (AS 3549) that outranks two Tier 1 ISPs.
|Org name:||Cogent Communications|
|Number of ASes:||2|
|Customer Cone size||17,484|
Cogent Communications is headquartered in the United States. Their portfolio has a steady stream of acquisitions (if you can’t beat em, buy em). They provide their services to corporates and other service providers.
Cogent’s network reaches some 34% of all IPv4 addresses, which is half that of Level 3. Their primary network (according to their map) seems to extend across America and Europe. It is presumed that the rest of the world, as for the others, is made possible by peering.
Global Telecom and Technology (AS3257)
|Org name:||Tinet SpA|
|Number of ASes:||1|
|Customer Cone size||15,623|
GTT exists here because of Tinet SpA – the Tiscali International Network, set up in 1998 in Italy. The international arm of Tiscali was sold in 2009, and was sold again in 2010 to Neutral Tandem, which merged their existing infrastructure with Tinet and renamed the business Inteliquent. Inteligquent’s data business and infrastructure were acquired by Global Telecom and Technology in 2013, making GTT the third-largest Tier 1 provider.
GTT’s customer cone covers some 4% of all ASes and 39% of IPV4 Addresses.
TeliaSonera International Carrier (AS1299)
|Org name:||TeliaSonera International Carrier|
|Number of ASes:||1|
|Customer Cone size||15,178|
TeliaSonera, headquartered in Stockholm, dominates telephone and mobile networks in Sweden and Finland and owns shares in operators all over Europe. In 2012, TeliaSonera came under fire for collaborating with certain dictatorships and allowing them to spy on citizens in their respective countries. They have also been investigated (in the same year) for bribery and money laundering in Uzbeckistan. As of December 2014, they have signed a deal to merge their Danish operation with Telenor, their primary competitor in that space.
TeliaSonera reaches some 33% of ASes and 36% of IPv4 Addresses.
NTT Communications (AS2914)
|Org name:||NTT America, Inc.|
|Number of ASes:||7|
|Customer Cone size||14,877|
NTT Communications is a subsidiary of the Japanese NTT Group. Unlike most of the companies listed here, it operates 7 Autonomous Systems and is comprised of 13 subsidiaries across the world. AS 2914 is the largest of these, spanning North America, Europe, Asia and Australia, and in size and rank eclipses the others. Taken together, NTT reaches 32% of ASes and 43% of IPv4 addresses.
Tata Communications (AS6453)
|Org name:||Tata Communications|
|Number of ASes:||2|
|Customer Cone size||10,230|
Tata Communications operates two ASes: one at rank 7 (AS6453) and the other at 1200. Taken as a whole, Tata reaches 22% of ASes and 28% of IPv4 Addresses. They operate the world’s first submarine “around the world” cable network ring and are one of the few Tier 1 providers targeting Africa and India. They have also made a recent investment in the Seabras-1 cable from the US to Sao Paulo, Brazil.
Telecom Italia Sparkle
|Org name:||TELECOM ITALIA SPARKLE S.p.A.|
|Number of ASes:||1|
|Customer Cone size||9,904|
TI Sparkle runs a global Internet backbone called Seabone, which links to every continent. The company itself is owned by Telecom Italia and keeps an exceptionally low profile compared to the rest. It covers some 21% of ASes and 18% of IPv4 prefixes.
Hurricane Electric, the next largest network, purchases transit. Cable&Wireless likewise; XO Communications, which comes next, bills itself as Tier 1 in the US but lack a truly global network.
“In February 2010, Telefónica CEO Cesar Alierta expressed in a meeting at Bilbao, Spain that his company intends to charge Google and other search engines for the use of their network. Alierta complained that such search engines were benefiting from the platform without contributing to the company’s expenses and that such a trend will change in the near future. Additionally he said that Telefónica will seek to push its own content.” – Wikipedia
Cesar Alierta’s statement here is only intended as an illustration.
Internet activists would like you to believe that the Internet is a free system, a democracy of hitherto unseen proportions.
Not really. As you can infer, we’re on the Internet because there’s money to be made out of us. This is not just a situation stretching to your local ISP – it’s how the very backbone of the Internet functions. The control wielded by these companies is vast – the 2007 case of Comcast blocking or slowing access for some subscribers is just a raindrop in the ocean. They’re capable of slowing down entire nations.
It’s also no surprise that the companies controlling the internet are essentially based in highly developed nations. This is as much historical as commercial, for it’s possible to correlate this with the formation of the Internet. Nevertheless, if must be point out that this also coincides with the locations of the companies, communities and individuals that produce or enable access to the most amount of content on the Internet. Whether by chance or by design, this sets up a stream in which these countries have a significantly greater chance of affecting thought processes worldwide. Good examples are the spread of memes and viral posts. Cecil the Lion became an international issue after Americans found fault with it. The Syrian refugee crisis was unknown until European media picked up on it. The Internet content giants are mostly US-based. This situation, however it arises, places the control of both data flow and ideas on the Internet in the hands of a few – but that is perhaps material for another study.
So what can be done?
Everything is purely commercial here, of course; these companies own the infrastructure they use to connect to the world. What I propose is a set of fail-safes to prevent countries from being solely at the mercy of corporations:
- Each government should outright buy, or otherwise set up, enough links to ensure that the government can remain connected to the Internet in case of corporate shifts. These lines would be used for government purposes only and not for private use or reselling. In the case of large continents with multiple countries, such as Africa, a Continental Union will be required to ensure this fail-safe.
- All operators should be annually scrutinized for quality of service relative to the infrastructure in place, to ensure that nobody’s pulling another Comcast.
- Developing regions with high populations should approach these providers directly for the construction of cable, with the simple promise that there’s a lot of people to be connected to here. Again, to use Africa as an example, if commercial viability can be demonstrated, it can go a long way towards enabling Internet access for citizens. It also allows the cutting out of middlemen.
- All countries should implement T&Cs of operation to ensure that the governments of the countries in which these companies are registered are not allowed to use this infrastructure, in however remote a way, to collect data, monitor, analyze or otherwise gain access to information that has not been explicitly granted to them by the customers.In the wake of the PRISM fiasco and subsequent revelations, I believe steps should be taken to prevent infringement on human and sovereign rights. After all, you never know when the man who sells you dinner might be forced to take a side.