This Research Bulletin has been published in Journal of Urban Technology, 11 (3), (2004), 1-34.
Please refer to the published version when quoting the paper.
The Internet is an urban technology. The telecommunications infrastructures which support it are predominantly deployed within and between major cities. In the last few years, extensive fibre backbone telecommunications networks have been rolled out across the globe creating a vast planetary infrastructure web on which the global economy has come to depend almost as much as physical transport networks (see 47; 21). Businesses and organisations increasingly reliant on high quality data communications demand choice and low costs for their telecommunications needs. As elsewhere, the availability of high bandwidth, high quality, low cost and multi-choice telecommunications services is of considerable importance to the economic development of territories across Europe. However, a two- or three-(or multi-) speed Europe of the digital economy has already emerged, in which, broadly speaking, core urban regions have significant advantages over peripheral rural regions in terms of access to telecommunications. This is unlikely to dissipate with the enlargement of the European Union to include much of eastern Europe, in spite of widespread policy backing and action towards a 'polycentric' form of territorial development at all levels in Europe (see 7; 8; 27). The uneven territorial development of telecommunications and Internet access is not then just another facet of the information society, but is absolutely crucial to understanding the socio-economic development of telecommunications as a whole, whether from a market, policy or academic perspective.
This serves as an overwhelming counter to those who suggested that the widespread development of telecommunications inevitably meant a diminishing importance for cities, geography and/or distance (see, for example, 43; 40; 39). Instead of this, more and more frequently, we find that the complex and intertwined relations between telecommunications and cities or geography actually reinforce the latter (see 15; 23; 53). Today, this is perhaps nowhere more evident than in the ways in which the telecommunications infrastructures deployed in and between major 'world' cities reinforce (in parallel with other types of infrastructure) the 'function' of these cities as "command points in the organisation of the world economy" (46, p. 3; see 44, for a comparison of Paris and London; 45), whilst simultaneously widening the gap with less globally-integrated cities and regions. However, the territorial development of communications (and other networked) infrastructures is far from being a fresh concern. Indeed, Gabriel Dupuy (10) argues that 'network urbanism' is as old as urbanism itself. The history of these infrastructures has been very much founded on differing approaches and strategies within and between territories. The development of the visual telegraph in post-revolutionary France, for example, was a means of symbolically 'reducing' the size of its territory to facilitate administrative and economic cohesion and control (29; 37).1 The subsequent lengthy periods of hegemonic state monopolies, and their focus on universal service provision, appeared to homogenise the intertwined relationship between telecommunications and territoriality. However, recent changes in the telecommunications market linked to demonopolisation and liberalisation on the one hand, and increasing technological sophistication and product proliferation on the other hand, are once again underlining the inherent territorial basis to telecommunications development, as companies focus differing services more and more often on very specific geographical or sectoral markets.
This paper offers an exploration of the ways in which the deployment of major backbone telecommunications networks in and between European cities over the past few years is bound up with territorial development in Europe as a whole. Building on research carried out under the programme of the European Spatial Planning Observation Network (ESPON),2 the paper takes its premise from similar recent work carried out in the United States, but underlines the crucially different geographical and regulatory contexts underpinning backbone development in Europe. The first part of the paper reflects upon and grounds the present research within emerging geographies of the Internet and its supporting infrastructures (part 2). Part 3 provides some background context to the development of backbone infrastructures across Europe in recent years, before part 4 goes on to present an exploration of the modalities and implications of these infrastructures from a European territorial perspective. This takes the form of both a broad analysis of key territorial trends present in maps of European backbone networks, and a more detailed network analysis of the Internet connectivity illustrated in the maps. In the final part, we offer some initial conclusions to this ongoing research.
2. INTERNET GEOGRAPHIES AND THE IMPORTANCE OF BACKBONE INFRASTRUCTURES
The Internet is frequently described as a global 'network of networks', comprised of many interconnected public and private infrastructures (19). Its origins in the late 1960s, as a US Defense Department project linking universities and military research establishments, and its subsequent development over the following thirty years, have been well recounted elsewhere (see, for example, 31; 1; 19; 51), but its rapid diffusion across the globe during the early to mid-1990s heralded a new round of global networking, which was intensified by telecommunications market liberalisation in many countries.
Whilst Internet access can be achieved via a number of different technologies or technical networks (dial-up modem through the basic copper pair, cable, ISDN, DSL, satellite, and mobile), the backbones supporting this access are hundreds of deployed terrestrial and subterranean fibre-optic cable networks.3 It has been suggested that the majority of telecommunications traffic now flows over these fibre networks, including the increasing level of digital data traffic which uses Internet protocol (IP) (35). Thus, these infrastructures can be seen as one of the dominant physical supports to the 'space of flows', which Castells argues to be the principal spatial form of the 'network society' (5). Indeed, the telecommunications circuits, networks of place-based nodes, and articulations of dominant interests that make up the three layers to the space of flows are all crucial and mutually constitutive elements to the development of these infrastructures.
Backbone infrastructures were thus constructed to allow the rapid diffusion of large amounts of information and data over long distances: "This places transit backbones at the top of the Internet 'food chain' as the networks that provide the transit services that make the Internet global" (19, p. 115-116). While individual users usually access the Internet primarily through either dial-up or digital subscriber lines (DSL) of the local telephone network or cable modem, larger users can often bypass these 'local' links and configure their own direct connection to the networks of global service providers. These 'fat pipes' thus serve the increasingly voluminous and bandwidth-intensive data communication needs of Internet Service Providers, technology companies and large corporations, which frequently demand their own leased lines and virtual private networks between widespread office locations. Indeed, these private networks are suggested to make up a greater proportion of the Internet than ordinary public networks (34). Decent bandwidth links are, then, crucial to most large companies:
Moss and Townsend also highlight how "the location and capacity of backbone networks and particularly interconnection points has a powerful effect upon the ability of firms in any metropolitan area to distribute large amounts of data and information via the Internet" (38, p. 8).
The concentration of this demand in cities has meant that carriers have focused their network roll-out strategies in inter-urban meshes. As a result, not only is the most bandwidth capacity present in and between major cities, but the most competition between providers is also there, resulting in lower prices and more tailored solutions. As Malecki has argued:
In this way, backbone providers have employed fairly explicit practices of 'spatial selectivity' - "a need to maintain hegemony, suppressing counterhegemonic interests and in the process attempting to gain, through pursuing a particular accumulation strategy, international competitiveness" (30). The inter-urban nature of this spatial selectivity in backbone telecommunications provision has a technical, as well as economic, basis:
The Internet is, then, primarily a 'network of metropolitan nodes' (6, p. 228).
Analysis of the geographies of Internet backbone coverage has begun to shed light on this 'urban bias' to providers' networks in a US context (52; 19; 51; 20; 34; 41; 35). As Malecki concludes: "To a large degree, the evolving infrastructure of the Internet is reinforcing old patterns of agglomeration: the world cities are alive and well" (34, p. 419). Within this, however, some of the research illustrates that 'connectivity agglomeration' (17) has appeared in a set of 'new network cities' (51), and that other cities further up the traditional urban hierarchy are not as well served by high bandwidth connections. For example, many of the above studies highlighted the fact that San Francisco, Washington and Dallas were better served by backbone infrastructure than New York and Los Angeles.
Network analysis of the topologies of backbone infrastructures has also been used to explore their territorial coverage and accessibility (52; 19). Through these studies, the crucial importance of the processes of interconnection or peering of networks with each other, which allows data and communications to be transferred between different networks, has become clear (see 36). Because of the generally high interconnection of backbone networks, and the subordination of distance to the amount of bandwidth between places, Internet traffic is not always routed along the most direct links. For example, before the construction of high bandwidth backbone networks in Europe, and the development of intense competition which brought prices down, it was not uncommon for Internet traffic between two European cities to pass via the USA, because it was quicker and / or cheaper (19).
With many of these networks tending to overlap in and between key urban centres, there is a good level of competition on most inter-city routes which not only creates a choice of provider and lower prices for clients, but enables Internet traffic to be routed faster between two nodes. As Gorman and Malecki outline:
The third implication of multiple networks covering the same routes concerns the importance of 'redundancy' for communications connections:
This is evidently particularly important for businesses that rely completely on the Internet for their operations, such as Internet Service Providers (35).4
Backbones and Territorial Economic Development
The vast sums of capital poured into the planning, construction and management of backbone networks in the last decade by dozens of companies (and the willingness of many of these companies to invest massively often with borrowed money and risk their futures on building these extensive networks) is symbolic of their strategic importance in the overall telecommunications market. Accessibility to existing and potential clients has been the main driver, as companies have sought particularly to offer increasingly seamless, end-to-end service provision to business customers located in major urban centres. Their own fibre backbone networks have created the capacity to be able to offer the most up-to-date services and by being routed directly between large cities allow businesses to exchange information (digital data) between their own office locations and with those of clients via the IP networks on which digital data relies for fast and efficient transmission. These telecommunications networks are, thus, the major highways of the information society, and remain the infrastructural foundation to the availability of competitive telecommunications services. What is far less clear, however, is whether offering connection to these networks creates any major kind of competitive advantage for particular urban places, and thus for companies located there. Finnie answered in the affirmative:
In addition, recent surveys by the likes of Healey and Baker5 show that companies place access to high quality telecommunications quite high up on their list of factors determining locational decisions,6 but we have little concrete evidence beyond this. During the dot.com boom, however, the importance of telecommunications access for property and property location were alleged to be high, as captured in the buzz phrase 'location, bandwidth, location' (26).7 For IT and media companies, access to several telecommunications network providers, redundancy to protect against service interruptions, bandwidth flexibility, and increased deployment for smaller companies (which are still big users) are major demands (25). Likewise, Gorman found a high correlation between e-business firm locations and bandwidth availability in the United States, suggesting that IP-based network infrastructure is viewed as a necessity for e-commerce (17).8 In addition, O'Kelly and Grubesic suggest that:
Indeed, some US research has explicitly attempted to explore the link between backbone network deployment and economic development in specific cases (42; 55 - cited in 35). Malecki's own statistical-based research indicated that bandwidth investment was higher in US cities with higher education establishments and 'economic dynamism' (35), although, of course, the causality behind this association could operate in both directions.
Indeed, given the predilection of major companies for obtaining the greatest choice, highest quality and lowest cost of telecommunications services, the location and extent of this type of infrastructure must have quite significant territorial implications for the economic development and relative competitive advantage of regions and urban centres (see 16, for an earlier analysis of these implications). Castells suggests that "Fibre-optic grids and advanced telecommunications systems have become a necessary condition for cities to compete in the global economy" (6, p. 239). It is therefore unlikely, for example, that a region without access to this infrastructure, without a 'glocal node' (24) or 'glocal scalar fix' (4), would be able to attract substantial economic investment, because major companies are unlikely to locate in such a region. The presence of multiple networks offers firms direct access to the globally integrated networks and services of the biggest operators, offering higher quality and more secure infrastructure, and faster data communications. Equally importantly, higher levels of competition in infrastructure provision (through the presence of multiple networks) are likely to lead to reduced telecommunications costs.
3. THE DEVELOPMENT OF BACKBONE NETWORKS IN EUROPE - SOME CONTEXT
Between around 1997 and 2001, the development of telecommunications infrastructures and services proliferated across Europe in response to the widespread liberalisation of EU telecommunications markets. One of the major trends during this period was a large-scale deployment of cross-border fibre-optic backbone networks to directly hook up the major urban centres of Europe.9 Prior to market liberalisation, the main telecommunications infrastructure had consisted of the national networks of incumbents (plus competitive networks in the UK following its early liberalisation policies)10, so this new networking period represented an 'upscaling' of the dominant geographical level of major telecommunications infrastructure in Europe (this subsequent bypassing of national boundaries constituted a significant difference to backbone development in the United States).
In the early stages, only a handful of operators initiated such pan-European strategies, but the business opportunities this presented to operators (connecting up only the most profitable zones of the largest cities in Europe), and the increasingly widespread availability of venture capital (see 56; 6) and substantial loans for them to undertake this vast infrastructural roll-out quickly led to dozens of operators (many of them new entrants) being present in the pan-European fibre market. As 'carrier's carriers', some focused solely on leasing fibre to their counterparts (particularly incumbents who were slow or reluctant to build their own networks) (2). Nevertheless, by largely basing their network deployment strategies on the European urban hierarchy, operators were both responding to and anticipating increased demand from major corporate clients, who dictated a need for more "seamless end-to-end performance" in (especially) data communications connections between their offices (9). The numerous fibre-optic networks connecting the main European financial centres, London, Paris, Frankfurt and Amsterdam, created a so-called 'Golden Square' of telecommunications infrastructures and services (12). This responded simply to the concentration of demand there, with the UK, France, Germany and the Netherlands accounting for 63% of pan-European traffic in 2000 (14).
However, by 2000-2001, these cities and the connections between them had become the main action scene for innumerable competing telecommunications operators to such an extent that there was a clear glut of fibre-optic capacity, which sent prices tumbling and forced smaller operators out of business, as in the case of Iaxis (12), and a glut of players - 'too many companies with too little to distinguish them and their business plans from each other' (54, p. 6). Malecki suggests that many of the new entrants in the global fibre backbone market were driven by an overspeculative 'build it and they will come' philosophy (35).
The substantial rethinking and restructuring of the European backbone telecommunications market in the last couple of years has consolidated the number of active companies (through mergers and acquisitions) and brought infrastructural investment to a complete halt. The vast numbers of parallel networks remain, but the proportion of international bandwidth being used on European routes has tumbled to around three per cent of total capacity, although this still represents many Gbps, or even Tbps, of communications (50).11 Indeed, the market shows signs of coming out of its slumber, rejuvenated by the fact that major demand for fast, seamless data and video communications still exists, and with constant technological advances and new market opportunities linked to the Internet, may increase in the near future.
In the next section, we provide a territorial perspective on recent backbone network coverage and capacity in Europe. Although in the absence of any completely up-to-date information on this subject, we are obliged to rely on 'pre-restructuring' data from late 2001, we suggest that this still permits a very useful exploration of the territorial trends in major backbone developments by highlighting where infrastructure is and is not present and to what extent.
4. HUBS, SPOKES AND CAPACITIES - TERRITORIAL PERSPECTIVES ON EUROPEAN BACKBONE NETWORK COVERAGE
Obtaining any kind of reliable data or detailed information on the extent and coverage of telecommunications networks is extremely difficult, particularly in such a competitive marketplace. Although some companies have viewed the publication of their network coverage on websites as a marketing tool, many others prefer not to divulge any such information, beyond the countries in which they are present, for fear of providing their competitors with even the slightest implicit insight into their geographical investment strategies. Nevertheless, resources can be obtained from consultancies which offer detailed overviews of telecommunications network coverage in particular zones. We use two such resources here, from the KMI Research and Telegeography consultancies, to explore the territorialities of pan-European backbone networks. The former allows us to analyse the broad 'hub and spoke' geographies of these networks from a territorial perspective, while the latter offers a further level of territorial detail in the shape of the total bandwidth capacities present along the 'spokes'. By focusing in parallel on network presence, inter-city connection and bandwidth capacity, our analysis begins to build up, for the first time, a detailed picture of the territorial development and implications of pan-European telecommunications infrastructures at the European level, regional level within Europe, national level, and intra-national level.
Source: http://www.kmiresearch.com/ (based on publicly available information or information shared with KMI as of Q3 2001). Copyright of map is with KMI, used by permission.
Figure 1 shows the map of 'Pan European Fiberoptic Network Routes Planned Or In Place' from the telecommunications consultancy KMI Research. This displays the extent of the infrastructures of 27 alternative (i.e. generally non-incumbent) pan-European telecommunications companies as of the 3rd quarter of 2001, and the cities interconnected by each network.
Figure 2: European terrestrial networks
Source: © PriMetrica, Inc., 2003 - http://www.telegeography.com/
Figure 2 shows the map of 'European terrestrial networks' from the Telegeography consultancy, which focuses upon the total lit, available bandwidth capacities installed along the backbone network routes of Europe. Analysing inter-urban bandwidth capacities allows us to explore the qualitative detail of the network connections in the KMI Research map. In other words, not just where the networks are and which cities they connect, but how important the overall links between European cities are in terms of bandwidth. In the absence of actual data on network packet flows, this type of information is held to be a good indicator of demand (35), as operators have clearly installed the most capacity on the parts of their networks where they had or anticipated the highest levels of take-up.
These maps are undoubtedly two of the most detailed available at a European level. The only limitation would appear to be that they do not reflect the downturn in the telecommunications sector a couple of years ago, which may mean that some of these networks were not finished or the fibre not lit, due to the disappearance of some companies and the consolidation or retrenching strategies of others. For example, Interoute is no longer operating at all, and Energis and Carrier 1 have cut back on their territorial strategies to focus on their traditional markets. Amongst other things, this underlines the difficulty in carrying out research in the constantly evolving telecommunications domain. Nevertheless, the maps do give a clear indication of the territorial pattern of private sector investment in telecommunications by (mainly) non-incumbents during a period when the market was 'working'. Indeed, as we are not so much concerned by the quantitative detail of pan-European backbones (eg exactly how much bandwidth is available on one route compared to another), for our purposes, the major territorial trends illustrated in the maps are unlikely to have changed substantially. The situation has also improved to some extent in recent months, which means that the slightly smaller number of consolidated carrier networks are trying once again to fill their pipes, even if it is estimated that only a small percentage of the bandwidth capacity present between cities is being used (50).
Territorial Analysis of Backbone Coverage
A number of important points of territorial analysis emerge from these maps. There is a broad 'three-level' core-intermediary region-periphery distinction at the European scale with the largest number of networks and the inter-city links with the highest bandwidth capacity most frequently focusing on a highly concentrated zone (the 'Golden Square' mentioned earlier) roughly delimited by London, Paris, the Ruhr and Hamburg (see figures 3 and 4 below).
Figure 3: Network connections in the European 'core'
Source: http://www.kmiresearch.com/ (based on publicly available information or information shared with KMI as of Q3 2001). Copyright of map is with KMI, used by permission.
Figure 4: Concentration of bandwidth capacity in the European 'core'
Source: © PriMetrica, Inc., 2003 - http://www.telegeography.com/
Thus, the most important connections (in terms of bandwidth) are to be found between the major urban (and business) centres of Europe. Table 1 lists the 12 main inter-city bandwidth routes in Europe with available capacity of more than 4.75 Gbps. The major trend is a German dominance with no fewer than 7 intra-German routes among the densest in Europe for bandwidth links. Given this, these inter-city connections tend to be short-haul routes as well, as telcos are evidently keen to maximise bandwidth between important, fairly proximate city regions, rather than deploy it along longer routes at greater cost and risk of remaining under-used.
Table 1: Major bandwidth routes in Europe (4.75-6.5 Gbps)
Source: © PriMetrica, Inc., 2003 - http://www.telegeography.com/
Nevertheless, in spite of the focus on a core area, if we look back more closely at figure 1, an oval-shaped patchwork web architecture of networks can be distinguished between Madrid and Stockholm, with a mixture of network corridors and in-between 'deserts'. It is also notable that the last four bandwidth routes in table 1 concern links beyond the core, extending high bandwidth capacity towards the north (Copenhagen), east (Berlin and Munich) and south (Marseille). This suggests that backbone providers were concerned with developing capacity along specific pathways in intermediary regions, which could then be further extended into more peripheral regions if necessary.
However, at the time the Telegeography and KMI maps were drawn, this was clearly not yet the case as only a small number of networks with relatively limited capacity extend to the periphery. It can clearly be seen that Greece, southern Italy, Portugal, Scotland, northern regions of the Nordic countries and eastern Europe (beyond Prague and Budapest) have little representation (see figures 5 and 6). Peripheral (and/or rural) areas of Europe would appear therefore to have relatively limited accessibility to the high-bandwidth networks of Internet backbones, and suffer from the 'end of track' phenomenon identified for Florida in US research (52; 41).
Figure 5: Limited backbone networks in eastern Europe
Source: http://www.kmiresearch.com/ (based on publicly available information or information shared with KMI as of Q3 2001). Copyright of map is with KMI, used by permission.
Figure 6: Limited backbone networks in southern Europe
Similar patterns emerge at the national scale, with the primacy of capital cities, such as London, and other large urban cities and regions, such as Milan, which host high level functions, set against the relative paucity of networks and bandwidth in peripheral regions such as the Highlands of Scotland or the Mezzogiorno of Italy. Germany would be an exception to this as its broad distribution of important cities across the national territory (Frankfurt, Dusseldorf, Cologne, Hamburg, Berlin.) ensures fuller coverage (see figure 7). This, of course, suggests that new fibre optic telecommunications networks reinforce existing spatial patterns as companies seek to address large scale markets hosting large corporate users.
At a lower territorial level, spatial differences in accessibility of pan-European telecommunications infrastructure become frequently very stark indeed, dependent upon the regional presence of an urban node on one or more of these networks to avoid being subject to a 'tunnel effect'. Gaping regional holes are left in the pan-European telecommunications 'web' by the deployment of networks along specific city-to-city infrastructure routes (eg motorways, railways etc) - for example, central France, central Sweden (see figure 8), and even central Germany (see figure 7). In addition, other long distance networks might pass through regions, but without connecting nodes, because they have been customised to link two particular cities and not the places in between - for example, the Energis network between Madrid and Stockholm appears to stop only at Frankfurt, and transatlantic networks being run from the USA into London do not connect cities in the south of Wales or western England. In this way, they are more like high speed trains or airline networks in terms of their network configurations than roads.
Figure 7: Backbone networks in Germany
Figure 8: Backbone networks in Sweden
In addition to these general trends of European territorial 'divide', however, we can also highlight some of the more divergent, and potentially more positive, trends which the maps show, based particularly around the notion of a polycentric form of territorial development of telecommunications.
Following the analysis of Peter Hall (27), the vision of a parallel and intertwined scalar polycentricity finds a very good illustration in the deployment of telecommunications infrastructures across Europe. It is clear, for example, that while pan-European telecommunications companies have traditionally viewed the 'global' cities of London and Paris as a crucial territorial foundation to their overall pan-European strategies, other cities and network links have become almost as important - the 'sub-global' centres of Hamburg, Dusseldorf and Amsterdam are more or less the equals of London and Paris in terms of network presence, and routes such as London-Amsterdam and Dusseldorf-Hamburg have similar bandwidth provision to London-Paris.
Operators were clearly also investing in cities outside the traditional European core, in a number of 'regional capitals' such as Madrid, Copenhagen and Vienna, which can be seen as the leading urban centres for telecommunications in part of the European territory. These, and other, cities (including Prague and Budapest) were presumably viewed as new or potential nodes capable of generating international traffic. Whilst not yet suggesting any "shake-up in the urban hierarchy" (34), these city regions might have the potential to become viewed as both 'new network cities' which surpass some traditionally larger city regions (51), and a crucial part of a more polycentric European urban system. In particular, some of these emerging urban centres may be viewed as 'gateway cities' for high-bandwidth backbone connections, in the way in which they act as links between the core area of Europe and more peripheral areas. Copenhagen does this for many of the pan-European networks which come from Germany and are destined for Scandinavia (see figure 9). Vienna and Prague have good network presence and quite large bandwidth connections because they act as 'gateways' between the core area of western Europe and the relatively new telecommunications markets of eastern Europe. Southern French cities such as Bordeaux and Montpellier must be passed through for those pan-European networks which have been deployed in Spain and Portugal. Similarly, in US research, the importance of cities such as Atlanta and Dallas on backbone infrastructure is held to be related to their 'gateway' geographies (52; 41). This illustrates the way in which bandwidth concentrates at 'funnel points' (34). This trend has already had important polycentricity implications because all these 'gateway cities' have become more crucial to the overall functioning and roll-out of pan-European telecommunications infrastructure than they would have been previously.
Figure 9: Copenhagen as a 'gateway city' for backbone networks
The development of a polycentric form of telecommunications territoriality at lower levels based around the 'spheres of influence' of large cities may be seen to be of two types. Firstly, the national territorial dominance of cities such as London and Paris has been such that telecommunications network deployment in the UK and France has been very much organised in relation to these cities. There is some limited evidence so far of national territorial polycentricity in telecommunications here - Birmingham, Manchester and Bristol are all increasingly important centres for telecommunications concentration, although still in the shadow of the capital, while Lyon, Strasbourg and Bordeaux have all profited from their 'gateway' locations (towards Italy and Switzerland, Germany, and Spain respectively) to improve their network presence and connectivities. Secondly, on a finer scale, smaller cities within the wider hinterlands of these key cities can be seen to have been able to participate in telecommunications network deployment, eg Reims and Rouen in the Bassin Parisien, and Reading and Cambridge around London, albeit largely through profiting from their proximate links to the capital city.
In countries without a real single dominant and influential large city such as Germany, a more tangible polycentric form of telecommunications territoriality has been able to develop. For example, we saw that many of the most important direct bandwidth connections in Europe are between German cities, and there are no less than five German cities with more than 15 alternative networks present (Hamburg, Dusseldorf, Frankfurt, Bremen and Munich). Both the overall centrality of these cities and their particular 'gateway' locations (eg Hamburg and Bremen link towards the Nordic countries, Munich towards eastern Europe, the Rhein-Ruhr cities towards France and the Benelux) are principal reasons for the promotion of this polycentricity.
Not all backbone networks in Europe have focused on the core area. Some more regional networks have concentrated explicitly on serving and connecting more peripheral cities, eg. Grapes and Silk Route in southern Italy and Greece, Infigate in eastern Europe, and Song (Telel) in the Nordic countries. Equally, other pan-European companies have combined the deployment of a very extensive network infrastructure with a series of particular regional or national network loops which link up a number of more peripheral cities to this overall infrastructure, eg. Telia in the Iberian peninsula, Energis in Poland (see figure 10), or WorldCom in Ireland.
Figure 10: The Polish 'loop' of the Energis backbone network
Analysis of Network Topologies
Another method of developing a territorial analysis of the backbone networks and capacities present in the KMI and Telegeography maps is to construct connectivity matrices of the nodes and links present on each network and the capacities along each inter-city route. This type of network analysis of Internet backbones has been done by Wheeler and O'Kelly (52), Gorman and Malecki (19) and Malecki and Gorman (36) for the United States. As Malecki and Gorman suggest: "Using a connectivity matrix for Internet analysis is even more suitable than for past transportation network analysis, since distance is essentially irrelevant" (36, p. 95). In the analysis that follows, we use the tripartite measure of Internet connectivity adopted by Malecki and Gorman (36) - binary connections, redundant connections, and bandwidth of backbone links - but in the first two instances, we prefer to focus on the total connections of a node via the networks it is present on, rather than just on direct connections, as we are interested in exploring the full territorial connectivity of European cities in relation to all other cities and not just the more limited measure of how many direct backbone links to other places cities have, which, in the highly concentrated urban core of Europe at least would poorly reflect the territorial connectivity implications of the inter-city 'corridor' architectures of many pan-European networks. The bandwidth measure, which is the most detailed of the three, based on the Telegeography map, will, however, allow some direct comparison between European connectivity and the US connectivity findings of Malecki and Gorman (36).
The matrix constructed from the KMI map contained 209 city nodes which were present on at least 1 of the 27 backbone networks illustrated on the map.
Figure 11: Comparing the number of pan-European backbone networks present in cities with the number of connections to other cities
Source: Based on KMI map, abstracted and plotted by the authors.
Based on a binary connectivity matrix, indicating the simple presence or absence of a connection (ie a shared backbone network) between two nodes, figure 11 plots the number of pan-European telecommunications networks present in European cities against the number of other places which a particular city is connected to via those networks (ie before taking into account the complexities of interconnection between different networks for which no details are available). As we would expect, the basic pattern is one generally characterised by the more networks present in a city, the more connections to other places that city will have. However, the gradient of the plotted points on the graph tends to even itself out as we move along the 'x' axis, which suggests that cities which are on more networks are only connected to a relatively smaller number of additional places compared to cities on fewer networks. In turn, this suggests firstly that there are a small number of very extensive pan-European networks which inter-link a large number of cities. This would explain how Gdansk has 119 connections to other cities by being on only 2 of the 27 networks, and Brno has 141 connections from only 3 networks. Both these cities are on the networks of Energis and Telia, and Brno is also on that of Carrier 1. At the same time, however, other peripheral cities on very few networks have far fewer inter-urban connections. For example, a Greek or southern Italian city present on 1 or 2 networks is thus only linked to 5 other places, eg Athens, Patrai, Naples and Bari. Meanwhile, other peripheral cities both in Poland (Bydgoszcz, Krakow, Rzeszow) and the 'Celtic fringe' (Dundalk, Cardiff, Aberdeen, Inverness) are also only present on 1 network, but that network connects them to 83 other places. We must clearly, therefore, distinguish between telecommunications networks in terms of connectivity and territorial extensiveness (in the first case, the Grapes and Silk Route networks serving Greece and southern Italy are very limited in extent compared to the Energis network serving Poland and the 'Celtic fringe'). As Malecki suggests: "Some locations are more productive or advantageous than others because they are also the locations of other networks" (34, p. 411). Although interconnection between different backbone networks will limit these territorial differences, because this interconnection takes place at specific nodes (cities with access to a large number of the networks), those peripheral cities already connected to many of these nodes via the network(s) they are present on will inevitably have faster, more efficient links to other networks (and other cities on those networks) than peripheral cities which are reliant on only one or two interconnection points. In this way, there is evidently more than one level or form of peripherality in European telecommunications territoriality.
Secondly, we can also suggest that beyond this small number of extensive networks, there is a larger number of networks which are either somewhat less extensive or simply replicate the routes followed by other networks. This would explain why being on the majority of the 27 networks featured on the KMI map does not lead to a city having many more inter-city connections. For example, while Hamburg and London appear on six or seven times more networks than Brno, they are linked to only 50 or 60 extra cities. In conclusion then, the density of networks in a city does not necessarily appear to closely correlate to significantly greater territorial connectivity on a wider scale. The differences between cities must therefore emerge in the quality and quantity of network connections between the same places, ie the number of networks offering the same route and the amount of overall bandwidth present on that route.
On a European level, however, through finding that there are no fewer than 45 European cities with links to 150 or more other nodes (see table 2 for the top cities), and only 35 of the 209 cities in our data set with links to fewer than 50 other nodes, we can conclude that while there are spatial differences between European cities in backbone access, Internet infrastructure is nevertheless very well distributed along the European urban hierarchy.
Table 2: The top ten European cities for backbone links to other nodes
Source: Based on KMI map, abstracted by the authors.
This simple measure of presence or absence of connection between nodes obviously does not tell us very much about the level of connection either in terms of number of networks or network capacity. Measuring the total number of all links on all networks between nodes gives us an indication of the cities connected by 'redundant' paths (ie two cities connected via more than one backbone). Core cities are interlinked by numerous parallel networks creating high numbers of redundant connections (see table 3), whereas peripheral cities with only a small network presence have relatively few 'extra' redundant links (and none if only one network is present).
We can note slight changes in the hierarchy between tables 2 and 3. For example, London is second for binary links to other cities, but only fourth when all redundant links to all cities are considered. Conversely, Dusseldorf, which was only sixth for binary links moves up to second in the table of redundant connections, while Paris moves from tenth to third, suggesting that the networks these two cities are present on have slightly more parallel inter-urban topologies (ie network routes which replicate other network routes). Another French city, Lyon, is present in both tables, and actually ranks higher than Paris for binary links, reflecting its 'gateway' location for southern Europe. In table 3, it is also interesting that Geneva and Zurich enter the ranking for redundant links.
Table 3: The top ten European cities for total redundant backbone links to other nodes
Source: Based on KMI map, abstracted by the authors.
The third measure of urban Internet connectivity concerns network capacities or the amount of bandwidth present along each of the links counted above. For this, we need to use the Telegeography map, which reduces the total data set to 146 cities with at least one direct path to another node of 0.5 Gbps available bandwidth.
Table 4 shows the top cities in Europe in terms of estimated total bandwidth of their backbone links to other nodes. There are quite clearly four top cities, London, Paris, Frankfurt and Dusseldorf, clustered closely at the top of the ranking, with Hamburg and Amsterdam making up a top six. We should note that although the KMI map showed it had fewer networks than Hamburg, London has nearly 20% more bandwidth on its links. Although other cities below these have still quite substantial bandwidth connections, we suggest that this is the only one of the three measures of Internet connectivity we have looked at where a clear division between a small set of top cities and the rest exists.
Table 4: Top European cities in estimated total bandwidth of Internet backbone links
Source: Based on Telegeography map, abstracted by the authors.
In the same way as Malecki and Gorman suggest that "there is a steady increase in the degree of concentration as we consider more comprehensive measures of Internet connectivity (ie from binary links to redundant links to bandwidth-weighted links)" (36, p. 101), we have found that a far higher level of concentration is present in the latter bandwidth measure than in the previous binary and redundant link measures. This is shown in figure 12. Internet connectivity remained quite high among numerous European cities along the urban hierarchy when measured in terms of binary links and redundant links, but when we looked at bandwidth connections, there was a set of top cities ahead of the rest, even if a number of other cities still had moderate levels of bandwidth too. Links to and from London alone account for nearly 5% of the total bandwidth in Europe, while the top five cities in bandwidth connections accumulate over one fifth of European bandwidth, and the top ten cities over one third.
Figure 12: Percentage of city totals accounted for by top-ranked cities
Source: Based on KMI and Telegeography maps, abstracted and plotted by the authors.
We can compare this important bandwidth measure to Malecki and Gorman's (36) US analysis, which also uncovered six top US cities for bandwidth connections. Where the crucial difference between the US and European findings lies, however, is in the fact that the top European cities do not concentrate nearly as much total bandwidth as do the top US cities (see figure 13). Figure 13 illustrates that, whereas the fourth ranked city in Europe still has over 95 per cent of the bandwidth of the leading city, for the US, the distribution has already fallen away with the fourth ranked city here having 73 per cent. While the amount of bandwidth then falls away quite steeply for Europe, it stabilises at around the seventh ranked city, after which it tails off more slowly. By contrast, the amount of bandwidth in the US continues to fall away quite rapidly, only stabilising towards the last few cities in the ranking. Indeed, it is notable that the 50th city in Europe has a higher percentage of the bandwidth of the leading city (London) than does the 23rd city in the US of its leading city (San Francisco). On this most revealing measure at least then, Internet connectivity has been deployed far more extensively down the urban hierarchy in Europe than in the US.
Figure 13: Rankings of European and US cities for total bandwidth as a percentage of the top city
Source: Based on Telegeography map and Malecki and Gorman (2001, table 5.2), abstracted and plotted by the authors
In conclusion then, using purely measures of which cities are connected by backbone links in Europe would suggest that Internet infrastructure is quite surprisingly very well distributed along the urban hierarchy with even relatively peripheral cities having some level of access. However, measuring European Internet connectivity in terms of the quality of these connections paints a much more concentrated picture, in which although there is still a large number of cities with quite substantial bandwidth links, a significant proportion of total bandwidth remains focused on a small set of core cities, reflecting our findings from the first half of this paper.
Building on research undertaken for the ESPON project, this paper has attempted to offer an exploration and some preliminary analysis of the recent development and territorial implications of backbone networks across Europe. Although some of the evidence from our research reinforces the findings of US research into the spatial aspects of backbone deployment, and its overwhelmingly metropolitan nodal qualities and focus, we have been able to identify that the European Internet infrastructure appears to be founded on a larger group of cities than the half dozen regularly cited for the United States. Although the two pre-eminent European 'world cities', London and Paris, concentrate a significant number of networks and amount of bandwidth capacity, other cities such as Amsterdam, Brussels, Lyon, Milan and four or five German cities remain highly crucial nodes, both for the backbone providers who were keen for a highly interconnected and widespread European network architecture, and for diffusing high bandwidth Internet accessibility throughout the European territory. In addition, beyond these 'core' cities, a number of 'gateway cities' such as Copenhagen, Vienna and Prague concentrate networks and bandwidth capacity which are funnelled into more peripheral regions of Europe (the Nordic countries and eastern Europe). At the end of the day, and even taking the recent market restructuring and consolidations into account, we can suggest that the major European Internet backbones 'rely on' a minimum of 12-15 cities to deliver high bandwidth networks and services across Europe.
If this is the case, we must begin to question why this appears to be a higher number of cities (or, at the very least, a similar number) than seems to emerge to serve the much larger US territory. Here, we invoke the major contextual difference between the US and European telecommunications markets - the continuing importance of national territorial specificities in the latter case. Market liberalisation in Europe may have removed some of the boundaries to competitive and trans-national network and service provision, but the particular institutional and political structures, regulatory practices, geographies and socio-economic environments of individual countries (and within countries, at regional and local levels, because these contexts and specificities are inherently multiscalar) remain very much in place (see 44; 11). Even the deployment of long-distance, trans-national Internet backbone networks between the 'world cities' of Europe is bound up in these territorialities. For a provider to deploy a point of presence in a European city, in order to serve its clients located there, all these intermingled factors must be taken into account (and in varying ways and extents in different cities). Territoriality will thus continue to shape telecommunications developments for some time to come, and in a European context at least, reinforce a need for multiple nodes at all levels (polycentricity) for the provision of access to high bandwidth (or lower bandwidth) networks and services. The differing spatial influences, qualities and implications of these territorialities for telecommunications development requires further attention, and will be the subject of our ongoing research.
* Jonathan Rutherford, Andrew Gillespie and Ranald Richardson, Centre for Urban and Regional Development Studies (CURDS), 4th Floor, Claremont Bridge, University of Newcastle, Newcastle upon Tyne, NE1 7RU, UK, Email: firstname.lastname@example.org, email@example.com, firstname.lastname@example.org
. In this way, telecommunications territoriality can be said to have emerged in Europe over fifty years before Samuel Morse demonstrated his later version of the telegraph in the United States.
2. ESPON is a programme under Interreg III, co-funded by the European Community and by the member states.
3. Wireless technologies are, of course, not excluded from this reliance on fibre backbones. Indeed, this reliance will augment substantially as wireless is used increasingly to transmit data communications (see 18).
4. The redundancy provided by multiple networks can be viewed, in part, as residual of the military origins of the Internet, when reducing the vulnerability of important communications networks to possible attack was seen as crucial (38).
5. For example, in Healey & Baker's annual European Cities Monitor, over the period 1997 to 2002, the 'quality of telecommunications' consistently appeared in the top 5 factors taken into account by companies when deciding on location (see 28 for latest figures).
6. For example, Grogan cites the chief operating officer of a management company, who argues that "today a company does not need to be high tech to require improved bandwidth - law firms require it, as do engineering, advertising, and a host of other professions. 'We look at it very simply. If you're in business and you're not communicating with customers directly, you will not be competitive in the 21st century, whether selling widgets, ads, computers, legal services, or whatever'" (26, p. 92).
7. As a result, one New York estate agent suggested that "if you're on top of an optic fibre line, the property is worth double what it might have been [...] Whose fibre (and what type of fibre for that matter) will be a major consideration in the site selection process. A perfectly built building in the wrong part of town will be a disaster" (3, p. 17; quoted in 22, p. 405; p. 408).
8. Nevertheless, in a recent study of the Internet adult industry, Zook (57) highlighted the ways in which places beyond traditional metropolitan centres are bound up in this particular form of e-commerce.
9. The first example of this was the Hermes network, which was rolled out along railway lines across Europe from 1996 onwards.
10. This meant that in order to provide international communications solutions, national operators leased lines off each other, but each part of the network remained under the management of separate operators, creating a fragmented, slow and expensive communications system, that left business users increasingly frustrated.
11. Indeed, during 2002, the Telegeography consultancy found that London had more than 6.5 terabits of lit capacity, representing approximately four times the total bandwidth requirements of the next 40 largest European cities (49).
Edited and posted on the web on 30th March 2004
Note: This Research Bulletin has been published in Journal of Urban Technology, 11 (3), (2004), 1-34