
Today’s post is written by Rudolf Van der Berg of the OECD’s Science, Technology and Industry Directorate
In 2012 the only submarine fibre optic cable that then connected Benin with global telecommunication networks and the Internet was cut for two weeks. International payments were not possible and the equivalent of 150,000 weekly salaries were not available in a country of 10 million people. The influence was particularly severe because most servers are located outside the country due to a lack of data centres and local-hosting facilities. Though similar cable cuts happen on average twice a week, their effects are generally less. This is due to the fact that most countries are connected to multiple submarine fibre-cables, connect overland to neighbouring countries, and have domestic data centres.
In OECD countries, networks look like a mesh with multiple paths that can act as each other’s backup. In developing countries, however, communication networks often resemble rivers, with small branches of regional networks delivering their traffic to a central national backbone that ends at one submarine fibre, making cable cuts a greater risk to the functioning of the economy.
A new OECD report – International cables, gateways, backhaul and IXPs – investigates developments in these networks and other essential components that are beyond a consumer’s “first mile” connection. Such networks are known as backhaul, backbone, regional, middle mile, core, trunk, or international networks. The report finds that there is still considerable investment in backbone and submarine fibre networks being made. The Baltic Sea submarine project between Finland and Germany, for example, aims to increase the operational reliability of networks in Finland, where currently traffic is routed via Sweden. In New Zealand, on the other hand, similar market initiatives to develop an alternative to the Southern Cross system, its only link to California and Australia, have not yet found the much greater commercial support that would be required.
Today, the decision to invest in a new cable can be due to a number of factors. It may, for example, be due to the need for shorter routes for high frequency trading. Here the few milliseconds gained in transmitting orders can result in a significant difference in the amounts of money earned (and in fact fibre-optic cables are too slow, so microwave networks are deployed between some stock exchanges). Nonetheless, all regions have to some extent benefited serendipitously from the initial over-investment in (inter-) regional networks between large cities during the dotcom bubble. Despite bankruptcies of the initial investors, the fibre is still present and has been bought and swapped, by telecommunication companies, cloud networks, Internet content providers and others. For example, Facebook and Google have both invested in submarine fibre projects and bought regional rings. By way of contrast, in other areas such as in many rural regions there is insufficient competition. Here, governments sometimes choose to regulate these monopolies to allow for competitive access.
Not all interventions in OECD countries, however, may be interpreted as stimulating the rollout of backhaul networks. In the United Kingdom, some believe the application of a “fibre tax” has a restrictive effect on deployment of backhaul networks. This property tax charges long distance network operators via a depreciating scale, based on the number of lit fibres (cables in use) that they have and on the length of those fibres, creating a competitive advantage for incumbents. It also requires operators to use more expensive equipment to employ multiple colours on a single fibre pair, instead of lighting unlit fibres.
It is not enough that countries are well connected through networks. The presence of data centres or other local facilities that can host Internet exchange points (IXPs) and servers is also essential. This allows local traffic to stay local. A new indicator of the number of websites under a country code top level domain name, a ccTLD such as .fr for France, hosted in the country, developed based on data provided by Pingdom, gives some insight into the functioning of the country’s hosting market. Just six OECD countries host more than half of their ccTLD domains outside that country, with Greece being the only country where two countries (Germany and the United States) host more of its ccTLD domains.
Name | Hosted in Country | Total Sites | Sites in Country | cctld |
Korea | 97% | 1750 | 1693 | .kr |
Germany | 92% | 25469 | 23306 | .de |
Japan | 91% | 14188 | 12964 | .jp |
Czech Republic | 90% | 4736 | 4258 | .cz |
Hungary | 84% | 2619 | 2197 | .hu |
Israel | 84% | 1302 | 1088 | .il |
Estonia | 83% | 456 | 378 | .ee |
Netherlands | 82% | 7937 | 6532 | .nl |
Turkey | 81% | 2095 | 1693 | .tr |
Slovenia | 78% | 518 | 403 | .si |
United States | 78% | 2709 | 2100 | .us |
Denmark | 76% | 2735 | 2072 | .dk |
Finland | 76% | 1444 | 1098 | .fi |
France | 74% | 10820 | 8021 | .fr |
Iceland | 72% | 193 | 138 | .is |
Norway | 72% | 1930 | 1389 | .no |
OECD | 72% | 164740 | 118479 | |
Poland | 71% | 14235 | 10176 | .pl |
Sweden | 71% | 3541 | 2522 | .se |
Italy | 68% | 8158 | 5513 | .it |
New Zealand | 66% | 1106 | 730 | .nz |
Australia | 65% | 7914 | 5140 | .au |
United Kingdom | 64% | 17532 | 11206 | .uk |
Switzerland | 62% | 2849 | 1755 | .ch |
Ireland | 60% | 1070 | 638 | .ie |
Slovakia | 60% | 1628 | 984 | .sk |
Chile | 58% | 1231 | 713 | .cl |
Portugal | 54% | 1347 | 729 | .pt |
Spain | 50% | 6129 | 3049 | .es |
Austria | 49% | 2614 | 1280 | .at |
Luxembourg | 49% | 174 | 85 | .lu |
Belgium | 46% | 2635 | 1213 | .be |
Canada | 45% | 4138 | 1873 | .ca |
Mexico | 22% | 3566 | 776 | .mx |
Greece | 19% | 3972 | 767 | .gr |
Econometric analysis of the extended global table, indicates that there is a strong positive correlation between the percentage of sites hosted in country and the reliability of a country’s energy network and ease of doing business in the country.
Local (blue) versus Foreign (Red) hosted content*
Source: OECD, Pingdom, Alexa
The new OECD report also looks at the issue of local infrastructure development and costs. The lack of locally hosted content sites makes it difficult for local IXPs to continue to develop, because a market for Internet traffic exchange (peering and transit) is weak. However, in some cases it is local players that refuse to exchange traffic locally. By not peering or buying transit locally, established networks force other ISPs and content providers to buy transit from them to reach customers in that country. If networks refuse to buy transit locally from these networks, but buy it from another transit provider, the traffic will be routed via an international link out of the country, to be exchanged elsewhere with the network. Forcing networks to peer (exchange traffic without settlement) has in the past not proven to be a successful solution. In exceptional circumstances an alternative could be to force networks to buy transit locally.
Useful links
For more on Internet traffic exchange (peering and transit):
A 2014 report on “Connected television”, explains peering and transit decisions between first-mile networks and content providers.
A 2013 report on “Internet traffic exchange” explains peering and transit and shows that out of 144,000 agreements 99,5% are based on a handshake.
* This map is included herein without prejudice to the status of or sovereignty over any territory, to the delimitation of international frontiers and boundaries and to the name of any territory, city or area.
3 comments to “Beyond the first mile: where your Internet comes from”
Iggy - 25/03/2015
Yes, forcing networks to peer has in the past not proven to be a successful solution.
Awesome article!