Any of these can help you achieve the goal which is to manage your costs by stopping your heaviest users from using as much as they like. Actually, ISPs do not need to do any more than that and this is the perverse nature of the problem: throttling demand is likely to lead to the most profitable model for service providers. It is perfectly understandable too because the way the market is structured: ISPs do not stand to gain from ever more, ever higher resolution bandwidth being available to their subscribers, do they...?

Traffic Shaping
Traffic Shaping allows you to discriminate the level of service (ie speed) offered to a user or a group of users at a given point in time for a given application. It is the most intelligent means of controlling traffic, because you can allow heavy users to “do their thing” when the network capacity is available. You do not have to “cut them off” when these users do hammer the network as you can reduce the capacity available to them at peak times without removing it altogether. This throttling can even be done at an application level.

A huge advantage of shaping is that it can differentiate between time sensitive, low volume traffic flows and “background”, high volume downloads (peer to peer applications in particular). You can configure profiles that take into account troughs in network utilisation and allow you to offer even heavy users unrestricted access at quieter periods, while controlling how much is used when the network is busiest.

Remember, it doesn’t matter to you if your customers are hammering their connections as long as that traffic doesn’t drive a higher peak on the network. There is a lot of free capacity particularly at night time and there is a certain balance in saying to your heaviest users, “do what you like as long as it doesn’t impact the rest of our customers.”

Traffic Shaping is the gold standard in traffic management because it gives you the ability to manage traffic for whatever group of users or applications that you consider to be a problem. This means that the majority of users (especially the 40% that use 3.8% of traffic in Ellacoya’s data) need never know that you are managing them. Just think of all those asterisks that could be removed from your product advertising…

The disadvantages of Traffic Shaping are its cost and the fact that the technology is being rapidly overcome technically by those who wish to continue old behaviour.

The cost issue: Packet Inspection is a very intensive activity and can add latency even if high end silicon is used. If you are committed to Traffic Shaping, you are committed to throwing ever more resource at the problem of recognising the users and applications that you want to shape and to installing enough hardware to achieve the results you want. A chain is only as strong as its weakest link and Traffic Shapers may well be your weakest link if you do not properly resource them.

One of the major problems with traffic shaping is that the technology is based on recognising protocol signatures. Heavy users don’t want to be traffic shaped of course, so those who wish to do so can stay one step ahead through encryption of their traffic or by disguising it as HTTP. Such “innovations” are designed to fool the shapers into allowing the files through unaffected and maintain the user’s download speeds.

Like a game of “whack-a-mole” played at ever increasing speeds it now looks unlikely that Traffic Shaping will ever be able to keep up with such innovations. The result is likely to be a fundamental changes from blacklist based systems (allow all except X), to whitelist based ones (only allow Y). If this happens, one of the side-effects will be the default shaping of any new protocols, even those that genuinely do not deserve it. Would Skype have taken off if the ISPs default policy had been to shape the application’s packets? How would Skype have convinced the ISPs to whitelist it?

Bandwidth Caps
Caps have allowed the introduction of ever cheaper broadband access since BT moved all of its ISP customers (I’m sorry, encouraged its ISP customers to move) onto Capacity Based Charging. This change in 2005 reduced the cost to the ISP of serving the majority of its customers, while maintaining the overall bill. Consider median usage around 1GB per month against average usage of 5GB.

A customer with a 2GB cap – more than is used by most – could be charged £18 (or even less) a month with the maximum exposure to the ISP fixed to around £11.70. Any incremental usage over 2GB is chargeable at rates varying by ISP. Mine charges me £1.49 per GB, which although it would not cover the full cost to them if I was to use it all at peak times (£2.34), is enough to make me think very carefully before firing up my Joost application.

So, on the surface, caps work for ISPs – if only they could get all those early adopters onto these not so new capped products they would be laughing – but there is a serious issue lurking in the shadows.

Bandwidth does not equate to value. People have a strong sense of the physical value of something and find it very hard to relate to the virtual cost of something like a DVD being 6.7 times (4,700 / 700) more expensive to transport than a CD.

You couldn’t blame a customer for calling your billing department saying “I’m not paying that much! I don’t have to pay any more to post a DVD than a CD, so why should I pay more to get it over the internet? I didn’t know it was going to cost that much. I won’t be doing this again, I can assure you…”

At current bandwidth prices, it would be cheaper to pay a taxi driver to pick up your Friday night movie from Blockbusters that it would to download it over the internet. This dynamic reminds me of the ill-fated ITV Digital purchase of First Division football rights – it would have been cheaper to pay everyone who ended up watching the matches to go to the games by taxi than it was to show it to them on TV. That story didn’t end well either…

Fair Usage Policies
This is the oldest and least confrontational of traffic management strategies, but also the least effective. Every contract has had an acceptable use clause in it, entrenched in the need to avoid legal exposure to criminal acts being carried out using the service. Over the last few years, these AUPs have morphed into Fair Use Policies that include some dark words about what will happen if your use of the service negatively impacts other customers.

Such clauses are often broadly written so that ISPs, not wanting to pin down what actually is a problem, can deal with one when they see it. The clauses don’t have the same deterrent effect as caps because they are not specific and do not have a price attached to them, but this vagueness is very marketing-friendly.

FUPs have the benefit that they allow the ISP to ignore irrelevant behaviour such as heavy off-peak usage or infrequent busy months for example, but the flip side is that it takes people-time to make judgments and enforce them which can be hampered significantly by organisational inertia. The cost of flexibility is often that nothing gets done.

Without a clear line in the sand, there is a risk that FUPs could lead the ISP into a rat-hole from which they cannot escape, where user expectations grow gradually until it is eventually impossible to arrest bandwidth growth without losing a large number of customers.

Fair Usage Policies often back up Traffic Shaping activity, so it is true that the above categories overlap but I hope that this gives a broad outline of what the industry is doing today to manage cost exposure.

Evolution
As stated earlier, such measures are really only in place to deal with the small minority, the top 5% who use far more than they pay for. The problems come though because of the mass market adoption of much higher bandwidth services discussed in article 1.

Today’s heavy users are a determined bunch. They know the ISPs that will act on FUP and they avoid capped products. Technology is allowing them to bypass the traffic shapers too, so the high-tech pirates can do what they do in a game of cat and mouse that probably adds to the fun of it all.

Tomorrow’s heavy users could be just regular customers as adoption grows of online video in particular. It has appeared for the last year or so that the content industry was going to go ahead regardless of ISPs and put its content up there, in many cases using peer to peer to avoid incurring any significant costs of their own. This to me showed a poor understanding of the market, ignoring the basic economics which said that this cannot be allowed to happen.

Traffic management is a self-fulfilling prophesy to some degree because of the uncertainty and deterrent effect it causes. The biggest winners are the owners of satellite and cable TV networks which will continue to satisfy demand for high definition broadcasts. Traffic Management will prevent the online video market evolving unless there are major structural changes, something we will begin to discuss next week.

Power
This is the final feature of Traffic Management. Perhaps the dynamic that will have the most lasting effect, long after usage caps are assigned to dustbin of history to sit alongside 0845 dial up access.

In 10 years time it is likely that we will look back on Traffic Management as being the first time that ISPs took control of what their users do with their services. Historically, ISPs have insulated themselves from what people do with their services by claiming to be “dumb-pipes”.

Now they need to control the traffic for their own financial survival, they will suddenly realise that this control can allow them to get much more from those who wish to use their distribution network to reach consumers. We may even see the pendulum swing for a while and find a situation where content owners feel “blackmailed” into agreeing with an industry that is suddenly vibrant and working together to the same set of objectives.

This control will be integral in solving the big problems on the internet today: I’m not just talking about video here, but piracy, pornography and criminal activities of all kinds. It is worth considering this can of worms because with great power comes great responsibility and traffic management does indeed give the ISPs great power.

Other Articles in the Series