Economics of Bandwidth

On 19th February 2013 I gave a short speech at the FTTH Conference in London on “Economics of Bandwidth” (in the Workshop “FTTH in the UK – A Detailed Analysis”). The text below is my speech.


I wish to make the case that the FTTH industry has an obsession with bandwidth that is blinding it to other equally important considerations. This vision defect comes from focusing solely on the central technology value of fibre. The opportunity is to widen that field of view to include peripheral advantages in market positioning. In this way, fibre can have a holistic economic model that generates both the highest benefits to users, and the best returns to network owners.

The nature of broadband demand and supply

Communications service providers are in the business of supplying a demand for doing a variety of useful (as well as useless) online activities. That demand is generated by computing devices, many of which can alone drive a 1 gigabit connection to saturation. The number of such devices in each home is expected to rise rapidly over the next ten years.

The supply on offer is a finite resource. Specifically, it is subject to two absolute constraints: a capacity constraint, and a schedulability constraint. The capacity constraint is the one we are most familiar with: every advert for ‘gigabits per second’ is promoting high capacity. The schedulability constraint is more subtle: given an offered load volume, it is possible to arrange its delivery so that each flow’s content arrives in an acceptably timely manner?

Many of the emerging high-value applications – such as small cells, home teleworking and online gaming – have tight scheduling demands. Such applications are sensitive to the flow impairment that must result when other users also contend for the network resource. That means multiple users of the same connection require strong isolation from each other in order for them to get a good quality of experience.

FTTH typically do not take explicit control over managing the schedulability constraints of different flows, and delivering their isolation requirements. A “best effort” approach is assumed to deliver sufficiently good outcomes.

Bandwidth is not a panacea

Offering more “bandwidth” directly raises the capacity constraint. A common misbelief is that more capacity solves these isolation issues. It does not: more capacity is not an adequate answer to the isolation problem resulting from the schedulability constraint. Indeed, under an increasing range of circumstances, adding bandwidth can paradoxically make it worse.

However, the FTTH industry been positioning high bandwidth delivered over fibre as the “final answer” to fixed-line user communications needs. That means it better have answers to the final questions on delivering good experiences, since users can readily reach both the capacity and schedulability constraints.

As a result, there is not, and never can be, a “too cheap to meter” effect for fibre.

Adding bandwidth to solve to schedulability and isolation issues means that the network has to be sized to the peak demand for both capacity and schedulability. As a result, despite the high innate resource capability of fibre technology, it will become capacity constrained. We have been here before: DSL was once seen as so fast, we were never going to have to worry.

The missing schedulability and flow isolation solution

The effect of reaching this limit will be a lack of flow isolation that causes applications to fail, and they will fail in an uncontrolled manner – unless the resource is managed. So where are your resource management solutions, so as to schedule demand in order to isolate flows, and ensure good quality of experience?

They are generally notable by their absence.

Unfortunately, more capacity is a very financially and politically expensive answer to a schedulability question. You have told investors and politicians that digging up the roads is a one-off, and no more capacity will be needed. Based on your current approach, more capacity will be required. You may add more wavelengths, but these truck rolls only get you capacity doublings; not hundred-fold improvements as you went from DSL. Those who don’t have ducts to pass more glass along will have to dig their way out of trouble.

As a result, there is a quality of experience hazard to users, and an unquantified cost risk to FTTH network operators. Savvy investors know that such risks exist, which is why they are wary of your promises of a bonanza from fibre.

A fork in the broadband economics road

We can continue to pretend we live in a world where capacity is the only constraint. In this world, the benefits that users had originally bought the fibre for can be lost. Your neighbour’s evening 3D TV downloads stop your femtocell from operating. Working from home is unreliable; every time the neighbourhood kids all come home from school, your teleworking systems fail. Your opex bills for lawyers will start to exceed those for lambdas.

With FTTH we’ve already used our one and only “cheap leap” in capacity: we’ve shot our one and only magic bullet. An unplanned demand for more capital will result in a loss of confidence in the industry. This has happened before, when we went from narrowband to broadband, and quality of experience failed to match expectations; the same happened when users were underwhelmed by the shift from 2G to 3G for wireless.

What we want instead is a world where there is a vibrant and rich range of services that users can access and depend on. In this better world, we have control over the network resources so that they offer predictable and reliable outcomes, both in terms of quality of experience and financial outlay. Only then will FTTH infrastructure be seen as the highly generative and valued asset it deserves to be.

The question therefore that needs to be addressed for a sustainable economic model is: how to deliver consistent service, given continually rising demand, and both capacity and schedulability constraints?

Engineer for outcomes, not just capacity

The answer begins with properly understanding these issues, so we can transcend the limitations of a bandwidth-obsessed marketplace, so as to experience the full upside of moving to fibre.

The first change is an attitudinal one: the industry needs to learn the value of understanding user demand, and delivering the service dependability that users crave. That dependability requirement creates an engineering challenge, which is to deliver predictable application outcomes.

Meeting that challenge means you need to learn how to isolate the flows, and assure the quality of experience for applications that users (and society in general) come to rely on. You need to learn about concepts like ‘stationarity’, and the inherent failure modes of existing network deployments, including broadband. With this information, you can plan to mitigate and manage them.

There are also business model changes that are necessary. The industry is currently based on ‘supply-push’ and ‘purpose-for-fitness’. This all needs to invert: broadband service providers instead need to understand users’ aspirations, and deliver fit-for-purpose services driven by a diversity of demand.

Furthermore, there are political changes that must happen. You must confront the network neutrality spectre. Being seen to be against network neutrality is like being seen to be against motherhood and apple pie – which is fine, until women can’t go out because they can only be mothers, and the consumption of any other fruit is a criminal offence.

With these technical, economic and political changes in place, you can construct the kinds of services that have both the quality and cost structure that users require. By raising the user benefits of the applications, fibre becomes truly fabulous, and the higher infrastructure costs become economically viable.

Please get in touch if you are looking for ways to improve the ‘fitness-for-purpose’ of your broadband network.

To keep up to date with the latest fresh thinking on telecommunication, please sign up for the Geddes newsletter