Connections +

Engineering & Design – The Bandwidth Gap

Opportunities in Access Network Environments - The rumours of a 'capacity glut' are false -- we are actually experiencing a 'bandwidth gap', argues David Prior of PBI Media. Closing this gap will provide an opportunity for access network providers and give impetus to the continued creation of core infrastructure.

September 1, 2001  

Print this page

Traditionally, network supply models have been derived from established models in the telecom environment. This does not work.

To quote Michael O’Dell of UUNET: “The deep intuition about network growth dynamics developed over years with voice networks simply does not yield workable results when applied to very large data networks.” The volatile, dynamic nature of traffic in Internets therefore provides for misrepresentation of demand and supply as derived from traditional models.

Figure 1 illustrates this misrepresentation. In the network model, the application of historical growth rates to demand forecasts provides for a smooth, flat rate growth figure. In reality, factors external to traditional forecasting, and unaccounted for in the growth model, impact the growth rate of demand and create step changes in the volume of demand experienced.

With many dynamic variables, demand must be considered using only those elements that can be considered static. In one technique, per capita user activity is disregarded with overall throughput per user elevated to the primary determinant of demand, and presented as offered load (demand offered to the network). In this technique, desired and actual throughput data is applied to previously determined data with regard to the numbers of ‘users’ and the proportion of simultaneous network audience.

The results of these calculations provide outputs relating to the Maximum Potential Demand (MPD) and the Access Limited Demand (ALD). Plotting the MPD against the ALD for a particular time period generates a chart that shows the gap between the resulting volumes of offered load.

Figure 2 indicates both the existence and the extent of what is called the “bandwidth gap”. The bandwidth gap exists because, as Herbert Simon inferred: “What information consumes is rather obvious: it consumes the attention of its recipients. Hence a wealth of information creates a poverty of attention.” With more and more information available, in more and more complex forms, the attention window gets smaller. The logical requirement from a user perspective is faster, more reliable, and more contextual delivery of information.

The economic model of supply and demand is, perhaps for the first time, driven by human desire as a result of the sociological aspects of the network. The impact on the economics of the network is one of inversion. “Supply curves slope down instead of up and demand curves slope up instead of down,” says Paul Krugman of MIT.

Coupling this effect with the sociology of a self-organizing network — that is itself subject to multiple dynamic influences — creates an environment in which the demand for improved access is relentless. This physical demand, in conjunction with load dynamics, creates a spiralling cycle of offered load. More importantly, from a provider perspective, this demand dramatically influences the nature of topological load in the network core.


In the paper “Scaling Issues on Internet Networks”, Bill St. Arnaud of CANARIE suggests that there are two key drivers causing growth in Internet models: increased usage and increased connections. These two items provide the realization that even if the number of nodes in a network remains static, the network load generated by those nodes will continue to increase as the capabilities of each node improve and the number of applications available to each node increases.

Figure 3 demonstrates the compound effect of logical nodes in the network. As more and more connections are established, by increasing numbers of both physical and logical nodes, the resulting ‘N-squared’ phenomenon begins to grow faster than the square of physical nodes. Determining provision in such an environment generates a model in which a small increase in usage, connections, or capacity in the access network dramatically increases the topological load at the network core.

Despite the existence of the bandwidth gap, the development of network topologies shows a distinct polarization as presented in Figure 4. The effects of this circumstance are twofold. In the first instance, at least two networks are being provided and maintained. In the second instance, the potential demand inherent in the access network remains limited by the access technology in use.

Results of research into the bandwidth gap have implied that ‘advances’ in access technologies utilizing legacy infrastructure were attractive as they ‘turned copper into gold’: extending the life of previous investment. In truth, what is required is the acceptance that traditional models have little place in a next-generation environment and that new models must be determined to support the transition from telecom to the next phase: communications.

On the cusp of this shift, any failure to ensure the full participation of the access network will reduce the rate of change and will continue to limit the potential inherent in the network as a whole. Providing the user population with the means to participate in the shift generates both positive and negative impact. The positive impact is that a step increase in operational functionality in the access network generates much greater demand in the core, allowing for greater utilization and revenues. The negative impact is that the cost of provisioning the infrastructure to meet that demand will also increase relative to some function related to the square of the connected nodes.


The ‘N-squared’ phenomenon also raises another issue: operators with smaller networks will be able to deliver a smaller footprint, high-capacity network at a lower cost than for larger networks. Recent events at very large-scale network providers indicate that this point is not only accurate, but also forms a guiding principle for the architecture that unlocks the bandwidth gap.

When you take these considerations into account, the idea that the next-generation network infrastructure requires a completely new approach gains ground. Although attempts will be made to deliver increased capacity and performance with traditional networking models, the implication is these will be both interim and short-term solutions. The network effects generated by the ‘N-squared’ phenomena in relation to Internet models will demand that any next-generation deployment be scalable, rational and proportionate.

The biggest question surrounding next-generation models relates to scaling. The assumption is that the model used must scale upwards, yet evidence suggests that a more viable approach lies in scaling downwards. This leads to the idea that the ISP peering process provides the foundation for a model that scales to a per customer level: a model in which each customer implements direct peering at a common upstream point. Building on advances in tuneable lasers, fiber delivery and functional convergence in the network, this approach is termed ‘customer empowered networking’ and represents a natural network approach to the distributed architecture of the information grid.

With the information grid, network topology comes to resemble a series of starburst networks, present on a fractal scale at community, provincial, national, regional and global levels. Provisioning this grid unlocks the bandwidth gap and ensures cost-effective, scalable and rational deployment of network services to customers. Leveraging the resulting economies of scale, this deployment serves to define areas of responsibility, focus and revenue generation for the provider community.


This approach requires a radical redefinition of what constitutes ‘the network’. The network is no longer delineated, but is instead a holistic model within which communications technologies are extended to the end user. It is only through a consideration of the environment as a whole that capacity will come to pervade the network and will begin to promote more and more advanced applications in the network with greater loads being offered to the network. These loads have the greatest beneficial impact and create the most highly-scaled opportunities for network players
at all levels.

Figure 6 illustrates the network change engendered by such a fundamental change to the network architecture. As Internet/IP reaches adolescence, it outgrows the capabilities of its surrogate parent — the traditional telephony network — and requires a topology that will support advances to the next-generation environment and beyond.

Dave Sincoskie of Telcordia Research has suggested that the reason the emphasis has been on the evolution of the core and edge network environments — to the detriment of the access network — lies in the scope of the deployment. As Sincoskie put it: “in the US, with 100 million households and average loop length of three to four km, there are 300 to 400 million km of loops. By contrast, with 25,000 km of fiber, you can have a pretty rich backbone network.”

In applying network considerations to this comment, it seems that the development and deployment of an optical solution to the access bottleneck in the US alone will require in excess of one thousand times the network route mile deployment that US backbone networks have demanded so far. In parallel, it would appear that the resulting opportunity from such a deployment is greater than that of the backbone market — and by a similar order of magnitude.

In Figure 6, it can be seen that the legacy infrastructure is expected to drop away. Presently deployed technology will require replacement, and there will be a need to deploy additional supporting capacity in the backbone networks relative to the access environment. These factors present a major opportunity for those who would provide optical capacity in the context of the access environment.


Closing the bandwidth gap will, however, generate additional impact in terms of topological load. At the end of this year, 87.34 per cent of potential demand is expected to remain locked in the bandwidth gap. Unleashing this potential demand as offered load to the network will generate a massive increase in the level of load experienced in network core, making all talk of capacity gluts redundant.

Figure 7 illustrates the forecast for demand locked in the bandwidth gap. This forecast is based on projections for users, simultaneous user population, and access technologies that are in use globally. What this image clearly illustrates is that without any action to reduce the bandwidth gap, the effects of the gap will continue to be felt beyond 2015.

What is needed is a reconsideration of the structure of the network model: the layers involved and the pressures exerted. Such consideration requires a redefinition in which the research, commercial, investment and political communities all play a part. The resulting model must derive from a common and co-operative perspective if the topological stress resulting from increased offered load is to be felt at the network’s core.

Closing the bandwidth gap ensures that topological stress in the network will increase. But resolving the bandwidth gap requires that new technologies are adopted, that evolution in the network becomes an accepted form of progression, and that new mindsets are formed to take advantage of the opportunities that such a resolution creates. The difficult part lies in persuading telecom investors and network operators that this investment is necessary, if not critical, to the continued evolution of the network as a whole.

At this time, such persuasion is doubly hard: stocks are falling, sentiment is highly negative, and beliefs are profoundly shaken. The impact of Internet and IP technology has not been felt as greatly as in the boardrooms of traditional telephony providers. These are providers who must address the need to extend service models, change the face and brand associations of their organizations, and diversify revenues in a time of ever-increasing competition and ever-decreasing traditional revenues.

The ability to think beyond traditional services and into a strategic position that is both highly diversified and highly targeted will become the most valuable asset of such companies. It is this mindset shift that will recognize the potential restrained by the bandwidth gap and that will permit the taking of steps that redresses that circumstance.


This cannot be achieved in isolation. Those who would generate access environments in specific locations must work with traditional providers to both reduce costs and accelerate deployment. Closing the bandwidth gap will unleash a flood of network load that makes today’s ‘demand’ experience look like the dripping of a tap in the face of a tsunami. New entrants will not be able to derive benefits from this flood on their own and must partner with those who have experience in managing customers, billing and providing more traditional communications services.

Without a sharing of expertise and experience, attempts to unlock the bandwidth gap will be weakened. Without a strong approach to the bandwidth gap, network offered load would increase slowly but not dramatically. Without increased offered load, the potential topological load will not be achieved. Without an increase in topological load and a commercial model that achieves real, definable revenues from the next-generation environment, market rumours of glut and negative sentiment in the telecom sector will continue.

Against such rumours, the present market climate will provide a beneficial outcome within which the emphasis on advanced connectivity, sustainable revenues, and customer-centric business practices combine to form the core operations of the next-generation network. Those who truly understand this model know that network load is a much more complex phenomenon than simple demand, and that the network itself is much more than just bandwidth.CS

David is Practice Director, eServices & infrastructure with PBI Media at the company’s London, UK office. Mr. Prior has held various information and communications technology roles in blue-chip companies in Europe, Scandinavia, North America, the Middle East and Australasia. He can be reached at

Print this page