Traffic sensitivity

There isn’t a discussion about internet pricing these days; it is a shouting match. Still, I am going to explain why I think usage based billing makes sense for wholesale access and as an option for retail internet services.

I will say at the outset that I am not engaged by any client on this file and I have no projects underway for any of the large facilities-based internet services providers (“ISPs”). The opinions expressed herein are solely mine.

An internet access network requires continual investment. I think all ISPs agree on this point. Traffic growth continues, driven mainly by increased volumes per user, now that household penetration rates have approached 80% (network traffic growth had formerly been driven by two factors: rapid growth in number of subscribers and growth in usage per subscriber as multi-media applications increase in popularity).

As a result, the dynamics of network engineering have changed over the past 5 years, to accommodate ever increasing traffic loads on the access and internet backbone networks.

Among a number of methods of reaching their subscribers, many alternate internet service providers make use of a Gateway Access Service (“GAS”) from Bell Canada, which aggregates customers from across Ontario and Quebec to deliver their traffic to the alternate ISP at a designated location. For example, an alternate ISP can connect in Toronto as the gateway to reach their subscribers throughout Ontario and Quebec. The connection from the subscriber’s home to the Bell local office is dedicated to the end-user, so there is no traffic sensitive component to that “last mile”.

However, for GAS, the ISP does not connect at the Bell local office; the traffic is aggregated with all other ISPs’ access traffic and carried to the point of network interconnection. This middle network – from the local switching centre to the point of network interconnection – is traffic sensitive. This is an important point that seems to have been lost in many of the most recent discussions.

Keep in mind, even with ‘per subscriber’ pricing, wholesale services were usage sensitive. Bell didn’t take the total network access revenue requirement and divide it evenly among the total number of ISPs. Completely doing away with usage sensitive pricing for GAS would mean taking the total number of ISPs that use the network and divide evenly. This is an approach used for some wholesale services that are not usage sensitive. That would have resulted in smaller ISPs paying more than their share, since Bell has by far the largest number of subscribers using that shared access network.

In its response to a CRTC interrogatory in the proceeding that led to Telecom Decision CRTC 2010-255, Yak set out its views on the right mechanism for usage sensitive billing [pdf, 69 KB, see Interrogatory 3]. It is a perspective that was shared by CRTC Commissioner Molnar in her dissenting opinion on Decision 2010-255 and it is echoed in my suggested Way Out blog post from a few weeks ago.

When the Gateway Access Service was initially designed, some kind of average traffic load must have been assumed. The original pricing for GAS was flat rated per subscriber; there was no component to take into account utilization. The service is a number of years old, so we know that whatever the utilization assumptions were at the time, they can no longer be appropriate for today. If the traffic assumptions from a few years ago fit today’s usage patterns, then it means that ISPs would have been overpaying all this time. When traffic growth on the network was being driven – at least in part – by subscriber growth, then increased ‘per subscriber’ revenues provided funding for reinforcing the access network. Now that subscriber growth has leveled off, the driver for network augmentation is more clearly seen to be growth in subscriber traffic.

Some would argue that only use in the busiest period of the day drives network investment. That is correct in a simple world of only a few subscribers. But, consider that access networks are actually part of a large national network. It is not uncommon for traffic between two points to be routed in a very circuitous manner in order to take advantage of spare capacity wherever it may exist. This is good network management. But at a certain point, additional network infrastructure gets added to deal with increases in traffic load. Recent traffic studies have shown that the peak traffic period is not much of a peak at all; depending on the user base, traffic in the busiest hour of the day is often only about 20% higher than the daily average, down significantly from previous years.

The peak load on one cross-section may get relieved by re-routing across routes that have a lighter load. But as average loads increase to be a higher share of the daily load, there are fewer options for routing around congested circuits. What this means is that more of each user’s total traffic contributes to the network engineering, not just the user’s load in the peak hour of the day. active network management and continual investment in more infrastructure by all stakeholders is a reason that most users don’t directly witness significant congestion in their service.

The pricing for network re-engineering can be dealt with in a few different ways. For wholesale access, the regulator could conduct a rate setting review process every few years to confirm that the engineering model used for wholesale pricing adheres to the current experience. That could theoretically be an adjustment to a flat rate price. Alternatively, the regulator can seek to align the rates to match the drivers of the cost. when the CRTC realigned the GAS prices, it adjusted the flat rate component downward and added a usage sensitive component. As I have explained elsewhere, I think the usage sensitive component was set incorrectly, but this does not mean that the principle was wrong.

The problem with a flat rate ‘per subscriber’ model is that it does not associate network costs with the driver of those costs. I don’t think heavy users – and by that I mean households that have usage that is more than the average – are bandwidth hogs. Knowing that traffic is increasing dramatically each year, I think that heavy users provide an interesting window offering insights on the behaviour of next year’s average subscriber. Each ISP needs to develop strategies on how to offer services to that segment, as well as have an understanding of the various services to offer other segments in the market. Consumers benefit from having a wide array of choices in the market – choice of service providers and choice in pricing models.

The wholesale pricing model needs to be fixed to ensure consumers continue to have those choices.

7 thoughts on “Traffic sensitivity”

  1. “I think the usage sensitive component was set incorrectly, but this does not mean that the principle was wrong.”

    This is really what this is all about though. Bell and the CRTC has made it very clear. It’s about making sure bandwidth hogs pay their fair share. It’s about stifling individual heavy users. If the Independent ISPs where charged for their total subscriber bandwidth (not usage) on a rolling period, I would be fine with this. This would be like charging for the use of the pipe based on how big of a share of the pipe an ISP needed at any given time. That would be fair, but this isn’t about fairness, no matter what Bell says.

    “I think that heavy users provide an interesting window offering insights on the behaviour of next year’s average subscriber.”

    If UBB, as it stands were to go through, this would not longer be true.

  2. Pingback: Tweets that mention Traffic sensitivity • Telecom Trends -- Topsy.com

  3. Hi Mark:

    Thanks for helping me understand this issue. I have a few questions, and I’d appreciate if you could shed some light on them.

    1. The Yak interrogatory pointed to by your article talks about “correlated” and “uncorrelated” usage. What are those?

    2. If wholesale GAS customers were to be charged based on the aggregate usage of their end-users, then what is the best way to set GAS usage caps and overage rates? For example, Yak suggested in their interrogatory that “the usage threshold should be based on the number of end-users multiplied by a monthly cap per end-user (based on Bells largest retail usage cap).” I personally feel that these usage thresholds and overage charges should be set independently of what Bell does with their retail customers. This way, the small ISPs will be able to completely decouple their business models from that of Bell. What do you think?

    Thanks.

    Vishal Malik

  4. I’ve been meaning to respond to this for the better part of a week, but I wanted to read up on some things before doing so. The post left me with a lot of questions and some of them I couldn’t answer. Mark, can you respond?

    “This middle network – from the local switching centre to the point of network interconnection – is traffic sensitive. This is an important point that seems to have been lost in many of the most recent discussions.”

    Wouldn’t a good solution to this problem be to allow ADSL-CO? That is, to let the ISPs handle the connection themselves the moment the wire hits Bell’s central office locations? Wouldn’t this let Bell wash its hands of all the so-called “heavy” users and make it the ISPs’ problem? They are already, for the most part, handling their own connection to the internet, leaving just this middle piece of the network as the problem point. Why, then, is Bell so clearly against ADSL-CO? Aren’t they insisting on having this network arrangement with the other ISPs’ customers? If so, this greatly reduces my sympathy for the network management issues that arise from it.

    “Completely doing away with usage sensitive pricing for GAS would mean taking the total number of ISPs that use the network and divide evenly.”

    Don’t we already have usage-sensitive pricing for GAS in the form of AHSSPI and EAS, which are charged on a per-GB basis? While the original GAS pricing was per-user, under the current situation don’t the ISPs actually *do* pay for that data on a per-GB basis? And wasn’t the wholesale UBB ruling actually going to add a new, additional form of usage-sensitive pricing? Finally, wouldn’t overturning the wholesale UBB ruling actually put us back to a status quo that *already includes* usage-sensitive pricing, and not, as you suggest, completely taking it away?

    “I think the usage sensitive component was set incorrectly, but this does not mean that the principle was wrong.”

    If the usage-sensitive component was set incorrectly, why not address that directly instead of adding another, additional usage-sensitive component in the form of UBB? And can you clarify on what basis you believe the component was set incorrectly?

    “Recent traffic studies have shown that the peak traffic period is not much of a peak at all; depending on the user base, traffic in the busiest hour of the day is often only about 20% higher than the daily average, down significantly from previous years.”

    Who, then, are these “heavy” users we keep hearing about and what makes them heavy? You can only pull data through the network as quickly as your bandwidth allows. If I have 10Mbps home service and you have 10Mbps home service, we can each suck data through the network at no more than 10Mbps at any given time. (I’m not going to get into traffic-shaping here – let’s assume that all the traffic you want to move is actually moving and you’re using the whole bandwidth.) That you and I are doing it at the same time is, according to your statement, not particularly significant because everybody’s pretty much moving a lot of data around the clock. So what difference does it make to the network if I downloaded 5GB of traffic last month and you downloaded 150GB of traffic last month? If, as you seem to be saying here, peak congestion isn’t really that much of an issue, what IS the issue? Total network load? How does UBB on wholesale address that in any way? The money question in all of this is “What problem are we solving with wholesale UBB?”

    “What this means is that more of each user’s total traffic contributes to the network engineering, not just the user’s load in the peak hour of the day.”

    I’m really not sure that is the correct conclusion. What it means is that the total load carried by the network is not varying as much based on time of day. If the spontaneous load is rising around the clock, fine. We can agree that lots of people are accessing the network all the time and maybe time-of-day pricing isn’t the answer if time-of-day usage is less relevant. But you make it sound as if the network having to move 150GB for you in a month is somehow more taxing than having to move 5GB for me in a month (ie “each user’s total traffic”). But networks don’t get bogged down from carrying a lot of data over time. They can either handle the current spontaneous load (which may not, as you say, be varying much based on time of day) or they can’t. It does not mean that total traffic moved over a month is more relevant than it used to be. It just means that capacity has to be increased because everybody’s moving more data all the time — how you determine how much capacity you need is still going to be based on concurrent spontaneous loads, not load over time.

    Even Bell’s Mirko Bibic, in his appearance on The Agenda, took great pains to point out that the issue with heavy network traffic is spontaneous demand. He used the analogy of trying to figure out how many trunk lines you need to handle everyone talking on their home phone lines at the same time, but there are many other ways to picture it. What floored me was that none of the other people on that Agenda discussion took Bibic to task for the fact that Bell has implemented UBB in a way that is not sensitive in any way to network load. They measure the total amount of data moved over time, with no regard to how much data people are moving *at the same time*. It therefore has no regard for how they need to size their networks to handle the load.

    Only by a very indirect effect of dissuading retail users from bothering to use their connection at all does UBB have any effect on network load. Many, including you Mark, have argued that UBB is not punitive but the punitive effect is the *only possible way* UBB as Bell and Rogers have implemented it could limit or drive down network loads. When Bibic was questioned at the Industry committee meeting why Bell did not implement UBB in a way that was sensitive to peak congestion, he could have said “there is no peak” if that point is correct, but instead he told MPs that it would be prohibitively expensive to implement time-of-day billing. How is that a good excuse to implement a billing model that doesn’t address spontaneous load, which is driving the cost of infrastructure upgrades?

    As much as I hate it, traffic shaping is the only reasonable alternative to expanding network capacity in order to address spontaneous load. I hate it because I don’t think it should be legal for Bell or anyone else to sell me a 10Mbps service and then bar me from actually reaching those speeds. If Bell’s network cannot handle me moving data at 10Mbps, they have no business selling me a 10Mbps connection.

    But that’s an argument for another day. We should get back to the cost of making infrastructure upgrades: As traffic on the AHSSPI/EAS/GAS service rises, aren’t the other ISPs buying more capacity, paying for more GBs of traffic to flow through those networks, and Bell is thereby taking in more money to pay for those infrastructure upgrades?

    An interesting comment was made by KvF when he appeared before MPs: he said that Bell is recouping its costs of running the wholesale network access through its existing fees. I’m 99% confident that’s what he said. If that is the case, why are we even talking about changing the fee structures in the first place? How is Bell successfully able to argue that somehow expanding traffic being borne by their networks on behalf of the other ISPs is unfair to them and forcing them to invest in infrastructure at tremendous cost if in fact the costs are being covered? I’m honestly asking.
    But I think I know why Bell is asking for it. KvF made another interesting comment during his appearance. He said that a sizeable portion of Bell’s high-traffic customers had shifted to the other ISPs, and therefore outside the reach of Bell’s retail pricing scheme including UBB. I believe he made this comment in order to suggest that somehow those customers who defected to the competition were doing the equivalent of moving their assets to an offshore bank and failing to pay their taxes. In other words, that by going to the other ISPs where UBB is not used, they were scofflaws. So much for ensuring competition – let’s treat people who go for a better offer from another company like cheaters! I think the real significance of his comment is this: Bell was getting a lot of money from those users until they switched to the competition. Now Bell is looking for a way to get those users back under its billing tent. They are just following the money. It’s nothing to do with “fairness” and that line about “regular” users “subsidizing” the “heavy” users (I take issue with all three quoted terms) is clearly born from a purely PR perspective. Who is Bell kidding? They are not in the fairness business. They are in the business of making money. I don’t begrudge Bell for trying to make more money. It’s a sensible thing for Bell to do.

    But you know who is supposed to be in the fairness business? The CRTC. I begrudge them for not thinking through the practical effect their ruling would have. The CRTC makes it clear they cannot and will not set retail pricing policies, but their ruling put those other ISPs in a position of having no credible alternative than to follow Bell’s retail pricing. If Bell wants to charge its customers that way, who am I to say they can’t? But applying Bell’s retail pricing model to the wholesale service was simple folly. IMHO.

  5. Jean-François Mezei

    The previous poster is correct. Those who support UBB convenienntly omit the fact that the ISPs already pay Bell Canada a traffic sensitive fee in the form of capacity based AHSSPI component of the GAS service. This is not fixed cost.

    An ISP catering to heavier users will have to purchase more AHSSPI capacity to serve those customers.

    And it is important to note that Phase II costing *garantees* that Bell gets paid not only for operating costs, but also infrastructure costs. So the whole argument that ISPs are getting a free ride is bogus because they help Bell pay for all those network upgrades.

    What is a crime is when Bell takes the money and instead of investing in the capacity needed by the ISPs, it invest in DPI equipment designed to slow traffic.

  6. This will be the discussion at the CRTC proceeding. I note that payment of a traffic sensitive component does not on its own mean that ISPs are already paying for all the traffic sensitive costs.

Comments are closed.

Scroll to Top