[AusNOG] News: Minister Conroy contemplating Government-Fundedundersea cable?

[Mark ZZZ Smith]

>________________________________
> From: Terry Sweetser (SkyMesh CTO) <terry+AusNOG at skymesh.net.au>
>To: ausnog at lists.ausnog.net 
>Sent: Tuesday, 2 October 2012 9:12 AM
>Subject: Re: [AusNOG] News: Minister Conroy contemplating Government-Fundedundersea cable?
> 
>
>I'm enjoying this debate immensely.
>
>And I don't think it is that off-topic.
>
>IMO, I don't think funding the NBN is a major issue for this list,
      given it's part "nation building" and part "social policy",
      funding is a political issue before it is a NOG issue.
>
>There are, however, NOG issues which also are very relevant to our
      hip pockets.
>
>Some things I have issue with:
>[0] The NBN is expensive at the wholesale level.
>[1] There are two many POIs.
>[2] CVC charges are not good value for 80%+ of the network.
>[3] The tail charges are too expensive at the 12M and 25M speed
      levels.
<snip>
>There's a major "not level playing field" paradigm with the POIs. 
      The ACCC has truly shafted every low cap and small rsp in the NBN
      wholesale market space.  The only telco with the lowest capex to
      reach every POI is the incumbent.  The philosophy of having a new
      monopoly was to break the incumbent's hold on the CAN: this has
      not happened yet, and it is now evident that the NBN will
      facilitate market domination for many years by the incumbent while
      the rest of the RSPs have to fund with real capex getting to these
      many POIs. That's [1].

So this seems to be fundamentally based on the assertion that the only successful RSP is a national RSP. Having worked for a very successful regional RSP, I don't think it is true. For example, that RSP has more than 90% of it's customers on it's own DSLAMs, in more than 30 telephone exchanges, and those telephone exchanges are all within an hour of where on-call (actually all) staff live. Minimising the number of wholesale layers there are between you and your customer helps reduce MTTRs. It can be much better to be a bigger fish in a smaller pond.

The other issue I have with the concern about the number of POIs is that financial costs aren't the only consideration - if they were, then the ideal number of POIs for the NBN would be a single one, so that an RSP would incur only one POI connection cost.

With the original POI model being 14, and the current one being 121, then cost is obviously not the consideration - availability is clearly one of the others.

Availability is a weakest link problem, and from an RSPs perspective, the weakest links in the original 14 POI model were the POIs themselves, regardless of how much redundancy exists within the NBN itself. If one of them as physically destroyed e.g. burnt down (perhaps intentionally), then at a minimum, your access to the rest of the population for services is now via the remaining POI.

For example, in the 14 POI model, the Adelaide POIs were not only the SA POIs, but also the NT POIs, and were also likely to be the intercap junction points between the east and west costs. If one of those POIs had been destroyed, would it be acceptable for the all SA, NT, and east-west services (numbered in the multiple millions) to go through a single POI? Then remember that to replace the other POI, even temporarily is likely to be at the very best probably a two months - is that an acceptable time to have a single point of failure for half of Australia's telecommunications services?. Also remember we aren't talking just Internet here - the NBN is intended to or will likely replace *every* other network - Internet subscriber access, EFTPOS and ATM, traffic/transport control, utility control, TV, mobile phone backhaul. All of those types of services will be impacted by a POI failure. The NBN and it's POIs needs to be as available as all the networks
 its replacing *combined*. Perhaps the only traffic that won't be on it is Department of Defence traffic - hopefully. Other factors that are being ignored in the 14 POI model is the expense of redundancy (i.e. with only two POIs, due to the traffic volumes you're likely to need high end and expensive equipment to carry it), and very little traffic locality (*all* traffic between subscribers in Darwin was going to be hair pinned to Adelaide an back).

I think the 121 model is far better than the 14 model, because it allows RSPs to be more selective about where they connect, the consequences of failure of a POI are far smaller, the costs of redundancy can be lower because their are opportunities to use more commodity components (1Gbps verses 10Gbps or 100Gbps interfaces e.g., and lower end platforms), and there is far better traffic locality. If you want to connect to all 121 POIs, your business should be large enough to sustain that. OTOH, if it can't, there are already wholesale providers who've said they'll be providing that sort of product, plus you can mix and match - directly connect to POIs within your budget, use wholesale providers to provide access to the others.

The numbers themselves are a bit arbitrary, however the smaller you go, the more consequence of failure, the less traffic locality, and the higher the technology costs of interconnecting. OTOH, a more diverse network tends to be a cheaper, more efficient and more robust one.

<snip>


Regards,
Mark.