[AusNOG] Mobile Data Capacity - Where's the bottleneck?

John Edwards jaedwards at gmail.com
Wed Apr 1 14:34:25 EST 2020

Hi Rob,

Looking at the Australian example, we can analyse the industry using the
ACMA database. There are about 42,733 mobile sites (devices with an ACMA
licence for Spectrum or PTS)  for Optus, Telstra and Vodafone. The same
database shows only 11,584 point to point microwave licences for the same
three carriers, including sites that aren't for mobile. So it's fairly safe
to assume that one quarter of mobile sites are already connected with
a wired service for backhaul that is almost certainly fibre.

Alternatively, if you're looking to build a new mobile network operator
from scratch - then yes the backhaul capex is significant but I'm sure
there are other carriers who will cut you a deal.


On Wed, 1 Apr 2020 at 10:18, Robert Haylock <robert.haylock at gmail.com>

> It's a nice idea, and I'm sure more efficient in Opex and over a long
> period of time with the benefits of upgrades like you say, but the Capex of
> deploying all that fibre would be huge, especially as cells get more
> abundant. That's why everyone really likes packets :)
> Rob
> On Tue, 31 Mar 2020 at 16:16, John Edwards <jaedwards at gmail.com> wrote:
>> Dark fibre to cell sites opens up more possibilities than just bandwidth.
>> Potentially the raw analogue waves from antennas can be re-modulated onto
>> DWDM wavelengths and then digitally [de]modulated in a datacenter.
>> This makes the whole process more efficient, reducing power and weight
>> requirements of components on-site, which means that existing towers can
>> carry more stuff. It also means that DSP resources that might be
>> intermittently used by a stadium or university campus can be re-allocated
>> somewhere else when they're not in use (there are a finite number of
>> customers on the network, after all). If the system needs a hardware update
>> to a new protocol, it only needs to happen at the datacentre, not by
>> ripping and replacing DSP's at the top of a 30M tower.
>> Once the industry is comfortable with that step, it can then start using
>> adjacent cell sites like MIMO antennas. No more need to roam to nearby
>> towers because you're already associated with them, no "hidden node"
>> problem, and spatial stream capabilities that allow for massive spectrum
>> re-use.
>> For added value, they might even modulate the same 20Mhz analogue signal
>> multiple times in the same wavelength. By slightly varying the phase of the
>> duplicated signals into an array of antennas, it might be possible to get
>> an electrical tilt effect in the antennas without any additional active
>> components on-site.
>> John
>> On Tue, 31 Mar 2020 at 11:33, Dave Taht <dave.taht at gmail.com> wrote:
>>> It is certainly my hope more will also deploy bufferbloat fighting
>>> solutions at various points.
>>> Typical cell bufferbloat is in the 1.6 second range, and would be
>>> worse if various protocols didn't just time out....
>>> _______________________________________________
>>> AusNOG mailing list
>>> AusNOG at lists.ausnog.net
>>> http://lists.ausnog.net/mailman/listinfo/ausnog
>> _______________________________________________
>> AusNOG mailing list
>> AusNOG at lists.ausnog.net
>> http://lists.ausnog.net/mailman/listinfo/ausnog
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.ausnog.net/pipermail/ausnog/attachments/20200401/9db3e32a/attachment.html>

More information about the AusNOG mailing list