November 30, 2011
Benoît Felten and Herman Wagter have published a follow up to their 2009 article “Is the ‘bandwidth hog’ a myth?.” The new article (for sale, but Benoit summarizes it on his blog) analyzes data from a mid-size North American ISP and confirms their original analysis: Data caps are at best a crude tool for targeting the users who most affect the amount of available bandwidth.
Read Benoît’s post for the details (or at least a fairly detailed overview of the details). But here’s the gist:
Benoît and Herman looked at the actual usage data in five minute increments of broadband customers sharing a single aggregation link. They looked both at the total number of megabytes being downloaded (= data consumption) and the number of megabits per second being used (= bandwidth usage).
They found that there is indeed a set of users who download a whole lot: “The top 1% of data consumers…account for 20% of the overall consumption.” But half of these “Very Heavy consumers” are doing so on plans that give them only 3Mbps, as opposed to the highest tier of this particular ISP, which is 6Mbps. So, even with their heavy consumption, their bandwidth usage is already limited. Further, if you look at who is using the most bandwidth during peak hours, 85.3% of the bandwidth is being used by those are not Very Heavy users.
Here’s the point. ISP assumes that Very Heavy users (= “data hogs” = “people who use the bandwidth they’re paying for”) are responsible for clogging the digital arteries. So, the ISPs measure data consumption in order to preserve bandwidth. But, according to Benoît and Herman’s data, the vast bulk of bandwidth during the times when bandwidth is scarce (= peak hours) is not taken up by the Very Heavy users. Thus, punishing people for downloading too much inhibits the wrong people. Data consumption is not a good measure of critical broadband usage.
Put differently: “42% of all customers (and nearly 48% of active customers) are amongst the top 10% of bandwidth users at one point or another during peak hours.” The problem therefore is not “data hogs.” It’s people going about their normal business of using the Net during the most convenient hours.
I asked Benoît (via email) what he thinks would be a more effective and fair way of limiting usage during peak hours, and he replied:
throttling everyone indiscriminately during actual peaks (ie. not predetermined times that could be considered peak) would be a fairer solution, although the cost of implementing that should be weighed against the cost of increasing the capacity in the aggregation, core and transit. The economics don’t necessarily work. And of course, that would affect all users, and might create dissatisfaction. But it would be fair and more effective.
In any case, the data suggest that “data hogs” are not the main culprits causing bandwidth scarcity. The real problem is you and me using our bandwidth non-hoggishly.
Date: November 30th, 2011 dw