[AusNOG] Death of the Unlimited Plan? [was] "Sewer broadband provider dumped" - Brisbane fibre plan scrapped

[Mark Smith]

Hi Bevan,

On Thu, 24 Mar 2011 01:05:32 +0000
Bevan Slattery <Bevan.Slattery at nextdc.com> wrote:

> Mark,
> 
> > I don't think that is byte charging. Byte charging occurs
> > when e.g. they give you 12Mbps and then tell you that you can't fill it
> > 24 x 7 with bytes (in case you want to), and that they'll also be
> > counting and billing for the bytes transferred. Are NBNco doing that?
> > i.e. is their monthly bill going to be something like
> 
> I think it is.  If you buy a ULL pair, you get full access to the spectrum/bandwidth available to the DSLAM port.  No CVC/AGVC/Byte (traffic variable charging) on the tail.  You can plug that straight into your backhaul (whatever you decide to use).  Today in a DSLAM environment  [Tail/ULL]+Backhaul
> 
> The NBN is [Tail+CVC/Byte] + Backhaul
> 

ok. The concerns in the article are then quite valid.

> It's like saying that buying bandwidth (IP transit) at $100/Mb/s/month isn't byte charging, but paying $0.01/MB downloaded is.

I think that is exactly true, $100/Mb/s/month isn't byte charging if
the bill doesn't vary based on bytes transmitted or received.

>  At the end of the day there is a direct correlation with bandwidth
>  consumption and cost.
> 

That is debatable. Are you talking Internet bandwidth or link
bandwidth?

Presuming your "bandwidth consumption" is referring to link utilisation,
I don't think there is much correlation. The total cost of two switches
plugged into the ends of a fibre link will have the same capital costs
regardless of whether the link is 0% or 100% utilised. Opex costs
are going to be pretty much the same as well regardless of utlisation -
the only variant I can think of is per-month electricity costs due to
link utilisation (e.g. creating the packets on an asynchronous link
like ethernet, they disappear on synchronous links like SDH/SONET), and
I doubt they'll vary much at all. If people care about accounting for
electricity costs, they do it on a per-device basis, regardless of the
types of attached links, their capacity and their utilisation.

I think byte counting is a way to try to account for network utilisation
costs across all the links the customer might use in the network,
because it isn't possible to easily account for periods of utilisation
on a per-customer, per-link basis. For example, it isn't easy to record
for a customer, for 1Gbps link A, you used 10% of link capacity for 3
seconds on Monday the 3rd of the month, for 10Gbps link B, you used 1%
of link capacity for 20 seconds on Tuesday the 18th of the month etc.
etc. and then bill for it. You could possibly achieve it by running
netflow on all your links, but who'd want to do that, and that's
obviously not practical when the network is the Internet. So coming up
with an average (or usually just picking a) cost per byte transferred,
and then counting the bytes at a single customer network point of
attachment is simpler and an actually achievable, albeit quite a
relatively inaccurate way of accounting for actual link utilisation
costs across all links a customer may use within the network. Domestic
bytes are cheaper than International ones because there are less
possible links domestic customers might utilise during a billing period.

Transit at e.g. 100$/Mb/s/month is a further simplification of that, for
which there is quite reasonably an additonal price premium. Flat
rate billing also has some benefits to the provider though, as
they now don't have to record byte counters, don't have to generate
varying bills and now have a more predictable and reliable revenue
stream.

As to whether NBNco need to employ byte counting, when they're
essentially providing virtual point-to-point links between customers
and ISPs, with strict and known traffic paths, is questionable.

Regards,
Mark.