[LINK] DNS outage?

[Paul Brooks]

Glen Turner wrote:
> On 29/07/09 08:33, Paul Brooks wrote:
>
>> This does depend on there being a different path for traffic to be
>> re-routed onto. In practice if you happen to be very close to the cause
>> of the break (or your intended traffic source/sink is), there may not be
>> another path to steer traffic onto.
>
> That's still a matter of money and technology. Is the last mile to the
> customer a non-redundant link or is it part of a loop?  As I posted
> before, I really hope the NBN adopts a loop topology for the last mile.
> That would go a long way to making the NBN more reliable than the current
> telephone network.
I wouldn't hold your breath, and I'm more ambivalent about the need.
I don't know of any serious contender that has a redundant loop all the 
way out to the residence. The current contenders for FTTN architecture  
- PONs of various flavours - are all non-redundant star networks, and 
have an effective range of ~30-50km or more. Compared to the current 
telephone network (a non-redundant star network from the exchange 
building up to 5 - 7km long (average ~2.2km)) and simplistically the 
non-redundant portion would appear to be worse, not better.

However, a PON OLT can generally service only up to 32 to 64 residences, 
so the length of non-redundant star fibre is likely to be limited to the 
radius that covers this number - in urban areas maybe only up to 500 
metres or so - so in urban areas, there may be an improvement over the 
current non-redundant portion.

Taking a step back though, the current last-mile network is very 
reliable - in my own experience, I can count on one hand with fingers to 
spare the number of times my connectivity has been down because of a 
last-mile failure. The vast majority of service impairments I've 
experienced have been due to problems further into the network, where 
loops, mulitple paths, multiple power supplies and redundancy measures 
should have kicked in transparently, but didn't.  Of course, we also 
have no knowledge of how many other failures deeper in the network occur 
where the redundancy measures prevent anyone  being impacted at all, as 
expected.

P.