[LINK] NBN to cost 24 times South Korea's faster network, says research body

Richard Chirgwin rchirgwin at ozemail.com.au
Thu Feb 10 15:05:29 AEDT 2011

<by mistake I replied to Marghanita but not to the list>

On 10/02/11 12:57 PM, Marghanita da Cruz wrote:
> Stilgherrian wrote:
>> On 10/02/2011, at 12:19 PM, Richard Chirgwin wrote:
>>> Define "aggregate"! :-)
>> Agreed. It's a bit like asking "Where does the traffic merge on the 
>> roads?"
> Well that analogy isn't too bad in this case, as I did mean the data 
> traffic.
>> In the NBN, traffic is aggregated in the very first length of fibre, 
>> because it's a herringbone-shaped GPON back to the exchange, or 
>> possibly a node a few streets away.
> The reason I raise this is that congestion/contention only seems to be 
> raised as an issue with wireless cells. There was extensive discussion 
> about the capacity of cells and prohibiting roaming between cells.
> There has to be some point at which congestion/contention becomes an 
> issue on the fibre.
Yes and no.

What do we need to experience congestion? Aggregate user traffic 
exceeding the upstream capacity at some point.

a. In the GPON part of the network I would consider severe congestion to 
be unlikely for most end users (will welcome correction).

b. After traffic is "handed off" to the Layer 3 provider - either the 
retail carrier or the intermediate wholesaler - contention is no longer 
an NBN issue. It's about the provisioning of the retailer's core network.

c. International traffic - contention will be that gap between aggregate 
user demand and "lit capacity" on the submarine fibres. Again, not an 
NBN issue. In my analyst discussions with ISPs over the years, I am told 
that the old "contention models" are less useful than they used to be. 
An ISP will look at either the average or peak demand on its 
international circuits, and will decide based on demand whether it needs 
to increase capacity. If total customer demand is less than the 
available capacity, then congestion isn't a problem.
> Richard,
> The example I had in mind was VoD/IPTV, Premium QoS Voice/Interactive 
> Video, which is ofcourse the compelling killer app, as opposed to 
> multicast cable TV.
> ie where did the problem occur with the afore discussed "via 
> broadband" ABC news reporter picture from Townsville. Was it in the 
> digitisation of the picture, the transmission capacity at some point, 
> or in the recording/decompression?
I've no idea. If, for example, the picture was being transmitted over 
3G, we have several factors to take into account.

- Distance from antenna (throughput falls with increasing distance)
- Cell loading (throughput falls with more users).

Initially I thought "few users outdoors, therefore low cell loading" - 
but if a lot of towers were knocked out, then cells may have been 
covering a large area, leading to high loading.

However, from other discussions, it seems to me that compression 
algorithms may have been the problem - the ABC was trying to put the 
pictures in a skinny pipe, so quality suffered.

I don't see this happening (say) in the NBN for IPTV content.

> Marghanita

More information about the Link mailing list