Broadband access to the Internet requires the seamless interconnection and coordination of many networks operated by different carriers. People refer to the Internet as a “network of networks,” because users simply “call up” web sites by keying in easily remembered domain names such as www.worldbank.org. Behind the scenes, a series of universally agreed to operating standards support the integration of the several links needed to connect consumers of content with sources. Carriers interconnect their telecommunications lines using common protocols that make it possible to identify and link users and sources of content. Many users also refer to Internet access as “cloud computing,” because the various interconnected networks appear invisible as though they operated in a cloud.
We can penetrate the obscurities of the Internet cloud to identify each and every network used for any particular link. Traceroute software and Internet web sites offer an easy way to generate a report on the specific networks used to reach a user-identified destination. To achieve global accessibility and connectivity, several different types of carriers participate. These carriers have transmission and routing equipment in identifiable locations, even though end-users generally do not think of the Internet in terms of the locations of specific devices and equipment. Internet access starts with a query or request generated by an individual, machine or sensor located in a specific location. A “retail” Internet Service Provider (“ISP”), which has installed wired or wireless facilities, provides the first leg of this Internet routing. The first link is often described as the first kilometer or mile, but the actual length may span vast distances via a satellite earth station instead of other local options such as DSL, cable modem, microwave and cellular radio.
The retail ISP connects upstream with one or more ISPs that may not provide service to residential users, or other “end users” including small and medium sized businesses. Many of these ISPs provide “wholesale” service in the sense that they concentrate on providing service to large volume users, such as multinational corporations and governments. Most of these ISPs own and operate lines with extremely large transmission capacity that provides long haul service across vast distances including transcontinental and transoceanic coverage. The ISPs may connect directly with content providers, or with a retail ISP that provides the so-called first and last kilometer service to and from the content source.
International connectivity among ISPs operates in a hierarchical fashion with comparatively many retail ISPs, and fewer ISPs that offer long haul services to subscribers in a diverse geographical area. Among long haul ISPs the hierarchy continues and becomes more exclusive with very few operating as so-called Tier-1 ISPs and more operating as Tier-2 ISPs.
Tier-1 ISPs represent the largest carriers having the most extensive and highest capacity networks. These carriers also carry the most traffic and typically qualify for interconnection with other carriers on a zero cost, “sender keep all” basis, i.e., no funds are transferred between two ISPs largely because a roughly equal amount of traffic originates from one carrier for onward delivery by the other carrier. The term peering* refers to the zero cost interconnection arrangements Tier-1 ISPs negotiate. Smaller ISPs typically secure access to Tier-1 carrier networks on a paid basis commonly termed transit, or paid peering.
As the Internet matures and more ISPs enter the marketplace, new hybrid arrangements have evolved that deviate from the peering/transiting dichotomy. For example, some Tier-1 ISPs have opted for “private peering” where they interconnect directly with another ISP, outside of Internet Exchanges.*
Some ISPs now agree to “multilateral peering” where more than two carriers interconnect on common terms and conditions at an Internet Exchange.
5.4.1 International Links
Fiber optic submarine cables and communications satellites provide the vast majority of broadband international links. For nations bordering on a coast submarine fiber optic cables provide the most cost-effective option as a cable installation can combine several cable strands, each capable of transmitting at a rate of several Gigabits per second (“Gbps”). For example the recently installed TAT-14 cable linking the United States with several points in Europe has a baseline capacity of 10 Gigabits per second that carriers subdivide into three service offerings: Synchronous Transport Module 1 (“STM-1”), approximately 155.520 Mbps; STM-4 , approximately 622.080 Mbps and STM-16, approximately 2.5 Gbps. The cable system has four fiber pairs configured for 47 channels each with about 10 Gbps in capacity, making the total design capacity of the cable system 3.2 Terabits per second.
Satellites cannot match fiber optic cables in the terms of overall capacity and bit transmission speed. However satellites can provide a cost-effective way to distribute broadband Internet traffic to many locations within a footprint as compared to the single point-to-point design for submarine cables. Satellites have the ability to transmit broadband Internet traffic at rates exceeding 15 Mbps, but more affordable retail offerings typically offer somewhat slower service, particularly for uploading content to the satellite.*
5.4.2 Internet Links
Having made the near complete conversion from analog to digital networks carriers have great flexibility in the manner in which they load traffic onto available transmission capacity. Technological convergence makes it possible for both domestic and international transmission facilities to combine voice and data traffic rather than use specific links for Internet traffic separate from conventional voice traffic lines. Instead, Internet links use transmission, switching and routing protocols optimized for data traffic, but also capable of handling voice traffic configured for transmission via Internet links. Put another way carriers no longer configure networks with an eye toward allocating transmission capacity for specific types of service. With traffic converted into packets of digital bits, both international and domestic links can handle bitstreams that subsequently will be converted into voice, data, video, text and other types of traffic.
Technological convergence means that a reference to Internet links has less to do with the nature of the traffic carried and more to do with the operating standards used by the carrier as well as the terms and conditions established for the complete delivery of the traffic.* This means that carriers are less likely to identify the nature of the traffic, e.g., telephone call versus video link, or to use legacy measures of the traffic, e.g., minutes of use. Instead the traffic will be identified in terms of the capacity and speed of the transmission link as well as the interconnection arrangements established for that link.
When carriers establish interconnection terms and conditions for Internet links, they increasingly refrain from applying the longstanding financial terms and conditions used in telecommunications and telephone service in particular. These legacy arrangements characterized interconnection as a “settlement” based on usage, such as voice minutes. Carriers handing off more traffic than they received from a specific carrier had to transfer funds to the “terminating” carrier. Carriers providing long distance telephone service negotiated a compensation arrangement, commonly referred to as an accounting rate, applicable to every minute of usage. For Internet links carriers are less apt to meter traffic in terms of time. Instead they will interconnect based on the capacity of the network links used and an estimate of overall traffic volume handled. For Internet links carriers substitute a measurement of minutes used with a determination of the bandwidth and bit transmission speed made available for the carriage of traffic originated by another carrier and routed onward to the final destination, or to the network of another carrier located closer to the final destination.
ISPs use different vocabulary and transmission measures when they interconnect Internet links. Also they generally use commercial negotiations to establish agreements, rather than rely on government forums, or regulated terms and conditions contained in a public contract known as a tariff. The largest Tier-1 ISPs, providing the longest links with the highest capacity, typically choose to interconnect directly with other similarly large and important carriers. Based on the assumption that Tier-1 ISPs typically have the same amount of transmission capacity available in different geographical regions, these carriers initiate interconnection negotiations with the expectation that they probably will not need to transfer funds. If two ISPs generate the same amount of traffic for each other to handle, then no money transfer should occur, because ISP A hands off to ISP B roughly the same volume of traffic that ISP B handed off to ISP A to handle. Such equivalency allows the carriers to “bill and keep” all funds generated from service. ISPs use the term peering to refer to interconnection arrangements based on traffic equivalency.
When two ISPs do not have roughly equivalent traffic volumes, the carrier generating more traffic that it receives incurs an obligation to compensate the other ISP. The term transit refers to negotiated terms and conditions when interconnected traffic volumes are unequal and a transfer of funds has to take place. The Internet links available to consumers combine peering and transiting capacity seamlessly so that in the vast majority of instances Internet access is available to any site, typically via more than one carrier and route.
Only in rare instances have ISPs refused to maintain an existing arrangement, or come up with acceptable replacement terms and conditions. When a dispute cannot get resolved two ISPs no longer will interconnect their Internet links. However, consumers usually do not experience service outages, because carriers typically negotiate several interconnection arrangements, covering two or more alternative routing arrangements, commonly referred as multi-homing. Except in instances where only one carrier provides “single homing” access to and from a content source and destination, one carrier’s decision to “de-peer” and not interconnect usually does not result in the inability to have Internet traffic routed to and from any source or recipient of content.
5.4.3 Implementation Issues for International Connectivity
Thanks to a common set of operating protocols carriers can interconnect their broadband networking with ease. For Internet traffic the Transmission Control Protocol (“TCP”) provides a widely used standard for traffic switching, routing and transmission. TCP helps support economic efficiency in the production of equipment, such as routers, by establishing a common standard useable by all manufacturers. While many devices are available and later vintages incorporate newly available features, the TCP supports high volume production of routers and other equipment produced based on the ability to sell them to all carriers and other users throughout the world. Put more simply the TCP establishes a standard “traffic cop” management function that most Internet equipment uses.
Additional enhancement of international connectivity results from the use of a common Internet addressing system, the Internet Protocol (“IP”) by both consumers and carriers. Having a common addressing system means that consumers need only remember the names attributed to desired Internet sites, e.g., worldbank.org. Carriers install devices, such as routers, that can read IP addresses and convert them to a larger sequence of numbers that corresponds to a specific installed device, e.g., the computer and the network used to originate a request for service as well as the name of the designated source of the requested service, or content.
The IP provides the basis for a universally supported addressing system that can establish order and promote ease in use by subscribers. While behind the scenes, ISP use routers to look up identities and locations of service requesters and providers from special servers containing such registrations, requesters only need to key in a single IP address. End users do not even have to know the procedures for assigning IP addresses and the method for organizing them. So called Top Level Domains refer to the type of organization housing the computer that originates a service request and the server that delivers requested content. For example .edu identifies an educational institution and follows its name or acronym. The complete address for the Pennsylvania State University in the United States combines www, to identify the part of the Internet providing Internet World Wide Web sites, the acronym psu followed by edu. Other top level domains include .com, .org, .mil., .co. and net representing in sequence commercial ventures, inter-governmental, multilateral and nongovernmental organizations, militaries, companies and networks. Additionally an IP address can specify the country location attributed to the IP address, e.g., www.bbc.co.uk for the United Kingdom based British Broadcasting Corporation.
While operating protocols promote international connectivity other factors have the potential to hamper progress and impose higher costs, particularly for users and operators least able to afford them. In order to achieve maximum international connectivity, the networks for each and every link in the pathway between content server and recipient needs to be optimal. Ideally each network should be in place and readily accessible either because one venture owns and operates each link, or because multiple ventures use common operating protocols and have agreed to mutually beneficial interconnection terms and conditions. In the multiple carrier scenario consumers can benefit by having access to more than one routing option as well as the potential that carriers will compete for both the long haul links and the first and last kilometer links.
On the other hand consumers may incur higher costs and inferior service if multiple carriers cannot easily interconnect their networks, or when few if any carriers are available to provide service. The weakest link—in terms of competitiveness and ease of interconnection—can have the most significant impact on the quality, convenience and affordability of service. For example, if the lack of network options forces a carrier to resort to indirect and circuitous traffic routing, both the carrier and its subscribers will suffer in terms of higher costs and inferior quality of service, including higher latency. The Internet’s operating protocols are designed to secure a complete link from the network capacity that is both immediately available to a specific ISP and represents one of the possibly many options for which the ISP has secured interconnection rights. If no such link exists, the ISP and the TCP/IP will search for alternative, indirect options.
ISPs and their subscribers in developing countries face the greatest possibility that they may not have access to the most direct and efficient routing, at the least cost. Direct network interconnection may not exist. The lack of competition might make a direct link prohibitively expensive. ISPs that might want to interconnect networks may lack an efficient and low cost way to do so. When direct interconnection cannot take place, indirect options have to suffice even though they add time, distance and cost to a link. For example, until nations in Africa had local or regional interconnection facilities, ISPs had to secure interconnection at distant facilities, some located many thousands of kilometers away. The term thromboning refers to the need to use circuitous routing to achieve interconnection that optimally could have occurred at a more convenient, nearby location.
Even for ISPs in relatively close proximity to each other, if direct interconnection cannot take place, a short link between countries would have to be replaced with two longer links to a third nation having ample interconnection facilities, e.g., nations in Europe. For example consider a still plausible scenario where African ISP-1, receiving a request to contact nearby African ISP-2 might not have a direct interconnection option. Instead African ISP-1 must route the traffic to a European ISP with which it has a transit agreement. African ISP-1 and its European transit service provider might interconnect at an Internet Exchange located in Europe. The African carrier would incur the costs to carry the traffic to Europe as well as the transit costs incurred when the transit providing European ISP routes the traffic to African ISP-2. The European ISP might provide the return path from Europe to Africa by itself, or via interconnection with one or more other ISPs with which it peers or transits, for onward carriage to African ISP-2’s network.
Years ago African ISPs might have identified ways to secure routing of their traffic via other generous carriers who might agree to provide transit services at low prices. The term hot potato routing refers to a strategy of handing off traffic as quickly as possible so that other carriers incur more of the total cost to secure a complete link. Because all ISPs now pay close attention to traffic volumes and the flow of traffic, ISPs cannot easily find inexpensive hot potato routing opportunities. ISPs bear the obligation to build or lease the facilities necessary to handle traffic without premature handoffs. However even if an ISP in a developing country did not want to handoff traffic to avoid network costs, having to do so now triggers higher routing costs.
ISPs not keen on exploiting hot potato routing opportunities need to have networks available to route local and regional traffic as well as the ability to interconnect with all needed networks. Because even the largest Tier-1 ISP does not own or least all the network capacity needed to reach any and all sites throughout the world, all ISPs need to have access to a facility that operates as a hub and interconnection point for many ISPs. These Internet Exchanges (examined in Section 5.5.2) provide the physical means for achieving local, regional and international connectivity by serving as the agreed to “meeting point” for all ISPs operating in the vicinity as well as those ISPs equipped to provide long haul transmission to other continents. Internet Exchanges provide seamless interconnection between local operators providing first and last kilometer service with other ISPs operating international links.