Pages

Pages

Sunday, May 09, 2010

Mobile broadband traffic - be careful about language

I am currently writing a Disruptive Analysis research report on mobile broadband traffic management strategies. I have discussed various concepts on this for the past year or so - the relative merits of offload, compression, policy management and so forth.

One important factor for vendors and operators to keep reminding themselves is about the importance of accurate language, logic and semantics. The wrong words can drive poor decision-making, especially on "emotive" issues. Non-sequiturs and logical fallacies can lead discussions or engagements astray.

One of the most mis-used words is "capacity".

What triggered this post was seeing a sentence along the lines of "3% of mobile data users take up 40% of capacity".

This is almost certainly untrue - as very few networks (none?) actually run at capacity-utilisation rate of above 40% - especially when averaged across all cells. If that were true, there would be almost-permanant and geographically-ubiquitous congestion for mobile data.

Add in to this the fact that "capacity" is actually an ill-defined term embracing multiple separate variables (uplink capacity, downlink capacity, signalling capacity etc) and measurable at various points in the network, and it becomes even more useless as a description of the current state of affairs.

What I expect may be the more accurate statement is "3% of mobile data users account for 40% of aggregate downstream traffic".

Which is an interesting observation - but not in itself a "problem statement", and certainly not something that can immediately lead to conclusions such as "... therefore flat-rate pricing is untenable" or "... therefore it is critical to manage specific applications".

Those are examples of non-sequiturs which are potentially damaging. There is no direct logical connection.

Instead, it is critical first to understand what the problem actually is. So, 3% of mobile data users account for 40% of aggregate downstream traffic - but what impact does that have, either on the other 97% of users, or the operator's cost base?

If that 40% of traffic was confined to rural cells operating at much higher rates in the middle of the night, it is likely that the impact on other users would be zero, although it might have some variable costs associated with peering. If that 40% was instead concentrated in the busiest urban cells in the middle of the day, when existing capacity really is creaking, then there's a much more pressing problem.

But what if heavy users tend to download a lot at night... but then have usage during daytime that is broadly on a par with everyone else? They are then not using capacity in a way that causes any more congestion than light users. It could even be that a nominally light user, doing a sudden big burst of mobile video at 9.30am on the bus to work, causes more problems than another user trickling P2P traffic throughout 24 hours.

And in each of these cases, there are varying signalling loads as well. A smartphone user checking his email 10 times an hour might be causing more headaches than a laptop user watching 15 mins of video once a day.

My view is that until there is really good, really granular data on actual usage patterns (and scenarios and forecasts for how that might change in future), knee-jerk comments about "bandwidth hogs" are likely to cause more trouble than they solve.

Instead, I am working on a priority list of actions that operators can take to reduce the pressures on the network without creating unintended consequences in terms of user experience, customer satisfaction, or fixing "the wrong problem".

There are various actions - and technological avenues - that can be pursued without risking money on over-complex solutions. I am particularly skeptical of policy management approaches that stress focus on application differentiation, rather than (for example) time-of-day.

Watch this space for more extracts from the analysis.

(As well as the research study, I am also sharing my views and data on this in private advisory consultations. Please contact me for further details - information AT disruptive-analysis DOT com)



NEW Mobile Broadband Traffic Management Paper

NEW Broadband Business Models Strategy Report

2 comments:

  1. The language I've seen was "x% of the subscribers generate yy% of traffic" (and not capacity).

    Regardless of the accuracy of either statements, operators do have congestion problems, mainly created by "dongle" (or broadband) users – and they invest in DPI/Traffic management equipment to solve it. They are not trying to fix something that is not broken - they think that limiting certain - applications (download and upload) will increase the number of users they can accommodate on the existing backhaul links without increasing churn.

    ReplyDelete
  2. DougV6:43 pm

    My view is that even if the statement is "3% of subscribers using 40% of aggregate downstream traffic" that they are fighting the last war. Its the small check-in app messages coming every minute of two that is really going to stress the signaling capacity of the networks ... even in LTE (but even more so with [W]CDMA.

    ReplyDelete