The Myth of the Bandwidth Hog Part 1

It seems like every method we use to access the Internet, wired or wireless, has some sort of data consumption cap. Some of those consumption caps are used to set a limit such that an additional revenue stream can be created at higher data volumes. But other caps are used to set a level that separates acceptable use of a network from abusive use of a network.

Whatever the language used or the commercial model implied, the existence of a cap seems to be fairly consistently aimed at the “top 1% of users”. It is said that the 1% consumes a disproportionate volume of data when compared to the average user. The implication being (without it really ever being stated) is that the 1% of users impose a disproportionate cost on the network operator. They are also commonly referred to as bandwidth hogs.

But there is a big problem with this. Bandwidth and volume are not the same. Bandwidth is typically measured in bits per second. Data volume is typically measured in bytes. To provide an analogy, a network’s bandwidth is equivalent to the number of lanes on a road. Data volume is equivalent to the number of miles an individual car travels along a road over a certain period of time.

The cost of building a road is almost fully correlated to the number of lanes that are built. But there is no correlation between the cost of building the road and the number of miles a car travels along it after it is built.

The same is true for a network. The cost of a network is almost completely determined by the bandwidth that is needed. And that “need” is extraordinarily predictable. Just as for a road system there is a network busy hour. The number of people you want to allow through a road system at a certain speed determines the number of lanes you build. And as a network operator, the number of people who you want to experience a certain bandwidth at the same time determines the total bandwidth you build.

Back to our 1% “bandwidth hogs”. Let’s say that this 1% all used their Internet connection fully for every minute of every day. If they did then they would consume a very large volume of data – lots of those bytes. Way more than the average person. But did they actually cost the network operator a disproportionate amount more? No. Imagine these bandwidth hogs actually turned their use off only during the busy hour – the hour that drives the cost of the network. Would removing 1% of cars from the rush hour on a road make any noticeable difference to the experience of the other 99%? No. Would those bandwidth hogs consume very much less data volume by refraining from using the network during the busy hour? No.

So there is no link between the 1% bandwidth hogs, when bytes are used to pick them out, and the cost to the network operator.

The problem is actually the large percentage of people who all want to use the network at the same time. That’s probably nearly all of us. And that’s a hard problem to solve. But making a scapegoat of people that continuously use an Internet service isn’t going to help. If all those bandwidth hogs left the network today the cost for the network operator doesn’t change and the experience we all have during the busy hour doesn’t get any better.

The following two tabs change content below.

Mark Taylor

I work as VP of Content and Media here at Level 3. English expat and passionate new tech energy evangelist.

Latest posts by Mark Taylor (see all)

15 thoughts on “The Myth of the Bandwidth Hog Part 1

  1. Very similar to the energy sector and building power plants. Although I would have used a different analogy than roads because the number of times driven on a road does effect cost because of wear and tear, vs data usage which doesn’t cause any wear and tear on network devices.

    • Thanks for the comment! That’s why I very deliberately said the cost of building the network. Yes roads have wear and tear but electronics also have a useful life as electronics degrade over time and need to be replaced.

  2. Another very nice article, thank you. However, I used to work for an ISP where a top % of customers had accounts that were running at pretty much the maximum theoretical level 24/7. As you state this is not much of an issue as the amount of potential bandwidth is based on the links / forwarding capacity etc of the network. The issue occurs when the ISP has to pay an intermediary for the data transferred on a central pipe, this is where single customer paying say $50 a month can cost the ISP hundreds or thousands of dollars a month in charges.

    Granted, for every user that abuses the service there are probably 100 or 1000 that use a very low amount per month.

    • Thanks for sharing. The same applies. If transit has to be purchased from an upstream network it is paid for on a bandwidth basis. The peak bandwidth in the month determines how much their invoice will be. So again it’s the total number of users online at the same time that drives the cost. Not an individual using the service 24×7.

      • Enjoy these types of articles. If the network is designed to handle peak, but typically used below peak capabilities, is the invoiced peak bandwidth amount referred to above based upon an amount that actually represents the full bandwidth used at the height of usage? Or is it based upon the lower average usage with an amount for the overage added to it? Similar to a 55mph road that fines (penalizes) anyone going 75mph.

        • That’s a very good question! And if we ever get to a pricing structure based on bandwidth that will be at the centre of the debate. I guess the rated speed of a consumer broadband could be lowered during the peak for standard users. If you want to run faster then you pay more. The question would then be for standard users is there a commitment to a minimum peak hour speed?

  3. Interesting food for thought, but I disagree with some of the analogy:
    1) Total data consumed is not analogous to miles driven. It is more analogous to how many times per day your car or your fleet of cars and trucks pass over a stretch of road or a bridge.
    2) A data hog isn’t like a single car on a freeway. Heavy data users are like expandable trucks that can grow to fill all the available lanes of the freeway simultaneously.
    3) The idea that turning off 1% of the top data users is equivalent to taking 1% of the cars of the highway during rush hour is worst case. This depends on actual usage, but most likely data hogs use a disproportionately high percent of the bandwidth during peak hours, too. Turning off 1% of the users during the peak hours may be like taking 30% of the cars off the road.

    Bottom Line: Setting data caps and overage charges is a fair way to limit unnecessary traffic and to help operators monetize the additional costs.

    • Hi Bill – The problem with your revised analogy is that with constrained upstream network the broadband consumer will use all their bandwidth for every communication. There is no local throttling. So someone sending an email will use all the speed for a small amount of time. Someone viewing a video will use the same speed (while filling their buffer) for much longer. So while not perfect (and I do realise that this is still a very simplistic view) I would still maintain that bandwidth, and not volume, are more highly correlated to the need to keep building out a network.

      • Mark, I think where we differ is our view of the correlation between data volume usage and peak bandwidth usage. When we started throttling our “data hogs” our peak traffic was reduced significantly.

  4. I agree with Bill which is the true anlogy. The 1% hoggers, limit the use of the other 99%, unless some kind of weighted priority is implemented.

  5. A version more colorful:
    Bandwidth and volume are not the same. Bandwidth is typically measured in bits per second. Data volume is typically measured in bytes. To provide an analogy, a network’s bandwidth is equivalent to the number of lanes on a road. Data volume is equivalent to the number of packages an individual vehicle traveling along a road can carry over a certain period of time.
    The biggest package carriers are the massive 18 wheelers. Imagine a five lane highway with nothing but 18 wheelers carrying all its packages to your doorstep. Next in size might be a 20ft “straight” truck as the trucking industry calls them. Then a 10ft U-Haul you might use to do an in-city move would be next. A pickup truck next. Then a minivan. Then a traditional passenger car. A motorcycle with a sidecar. Then a scooter. Finally, a bicycle messenger. The bicycle could be a mountain bike on a gravel road with one package in the messenger’s bag. That last example would be like a POTS line. If you have an 18-wheeler’s worth of data (a 4GB DVD) and a single 56kbps POTS line, you’re going to be waiting a very, very, very, very long time to watch Thelma & Louise drive off that cliff.

  6. While your analysis is technically true, in actual fact I expect most users use their Internet connections most during prime time. It’s most convenient for them (that’s WHY it’s prime time), and there’s little incentive for them to shift to other times (I’m old enough to remember ISPs and timesharing services that charged less during off-peak hours — but that has mostly disappeared). This is especially true now that most bandwidth is being used for real-time streaming — you can’t let this download in the middle of the night, because you have to watch it live.

    So total volume and bandwidth use are likely to be very closely correlated, and the former can be used as a proxy for the latter. And total volume is easier for users to track and control than network congestion that the providers really care about. It’s an approximation, but a useful one.

    • Barry I agree. My point is should the consumer broadband industry, in order to generate extra revenue to pay for network expansion, target people who use the network all day and consume lots of bits? Or should they seek an additional revenue stream from people who use it during the time that is driving the need to continue to add capacity to the network? That might be tough because it is most of us, as you say. But I also think there is then scope for some application innovation that may limit peak demand. Plus, as I said in the second part, maybe a standard broadband package has a throttled rate during the busy hours with a premium package providing an unthrottled rate.

  7. Pingback: The Myth of the Bandwidth Hog Part 2 - Beyond Bandwidth

Leave a Reply

Your email address will not be published. Required fields are marked *


4 − three =

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <strike> <strong>