Ultra Fast Broadband (UFB) is here if you live in Whangarei, where local lines company NorthPower started building a gigabit passive optical network (gPON) in 2008 and recently signed with Crown Fibre Holdings (CFH) to be the area’s Local Fibre Company (LFC). Circuit speeds start at 30mbps download / 10mbps upload, with a Committed Information Rate (CIR) of 2.5mbps symmetric (up/down). While an end user may get 30mbps (or higher for some plans) sometimes, CIR is what they’re guaranteed to get at all times between their homes and the local exchange. Once traffic gets to the exchange and is handed over to an Internet Service Provider (ISP), Committed Information Rates are unlikely to be preserved. It’s simply too expensive for a regional ISP to preserve CIR . The diagram below shows a minimal one-POP ISP networks with 1000 subscribers, preserving 2.5mbps CIR from the subscriber to the Internet. It makes the assumption that 80% of traffic will be national and 20% will be international. This assumption takes into account the availability of various Content Delivery Network caches on both peered and unpeered national networks.
The network has no redundancy. If any one element or circuit breaks, service will go down for all subscribers. It assumes the minimum possible equipment specifications for the amount of subscribers and traffic. It does not include web hosting, email provision, or in-network caching. An operational assumption is that one spare unit will be kept in the ISP office for each active element – a “cold spare”. The table below explores the monthly costs involved in maintaining our example ISP. The pricing assumes that for negotiable costs, such as backhaul, transit, and equipment finance, the best possible discount is achieved for a company of this size. It assumes contracts start today, and that the ISP is not locked in to existing long-term contracts which could be at several multiples of current rate card pricing. It assumes a manager, an engineer, and a junior employee on helpdesk. It assumes both the manager and engineer will cover helpdesk as necessary, and that all three employees will work at below-market rates. 
It is clear that in order to provide 2.5mbps CIR to the Internet, assuming that 80% of traffic remains domestic, a regional ISP would need to charge $154/month per subscriber in order to break even. Even then, the three staff would have to rotate on-call duties to satisfy help-desk and upstream transit provider requirements of 24 hour reachability. Maintaining the 2.5mbps CIR (already a 20:1 oversubscription of the 50mbps peak rate) would break the business.
Interesting breakdown.
It seems pretty crazy that national un-peered is almost as expensive as international.
Even if you dropped to 500Mbit national un-peered and 200Mbit international, you cut what 20-30k of cost not enough to get to <$100.
Wonder how UberGroup are able to offer those $99 plans.
LikeLike
The implication being contention on their international and national un-peered is awful.
LikeLike
This is very useful Jon.
The opex for your scenario is likely higher than you have outlined. It would be good to see subscriber acquisition costs and churn impact. For example, the RSP will ahve to pay sales and marketing costs to acquire the customer. This can be as high as $100 per acquisition (in fact likely higher, given that with the above staffing scenario all of this will need to be outsourced). To maintain the customer base at 1,000 subs, you’ll need to factor in the cost of churn. To acquire new subscribers, your example RSP may need to subsidise the early termination charges (ETCs) of competing RSPs. As with the mobile world, these ETCs are often punative and do not just reflect the residual value of a piece of amortised CPE. They can be quite high and lead to a long payback.
LikeLike
I can take $52K/month out of your opex straight away as an established independent ISP and no need for regional backhaul. That drops your price to $105/month for break-even which is close to the predicted market rate of $100/month that I’ve been banging on about s the top of market for ages now. Acquisition is a valid strategy.
Other than that, I can’t see anything wrong with your estimates and the overall analysis. I do not see any way to make money from UFB at this stage. It requires national backhaul to drop to <$10/month and international <$30/month to confidently break even. That's 2014 at the earliest. I welcome our new UFB overlords as the first market entrants and await the liquidation sales.
LikeLike
So if you’ve got an established business and you enter UFB trying to preserve a 2.5mbps CIR to the Internet, you’re going to lose money on every new subscriber. Sounds great, sign me up!
LikeLike
As mentioned previously, I am in no screaming rush to enter the UFB market.
LikeLike
I am quite.. intimately aware of dom vs. intl split in some large NZ networks, who have on-net CDNs (Google, Akamai, etc.)
The intl is more like 80%, assuming on-net CDN is counted as domestic, not the 20% you’ve assumed here.
LikeLike
Oh, well that’d make the economics of maintaining CIRs significantly worse.
LikeLike
CIR intra network, let alone International, what a fantastic way to waste resource in fixed allocation, that will rarely or never be used.
The CIR concept is useful in the autonomous network, not beyond. It was useful when one operator owned the switch, but in a collaborative network, it’s a useful internal metric and customer comparison, nothing more.
Thus it’s hardly surprising that if you take on a Quixotic pursuit like sustaining a broadband CIR beyond it’s useful handover point you will get:
a) No improvement to a lot of sites
b) Ludicrous costs
Taking the CIR international is a beau geste that I suspect would leave a significant proportion of that expensive resource languishing at the far end due to weaknesses in other networks and services.
What would be an interesting comparison would be between this model (1K, 20:1, 2.5Mbs CIR) and the current UBA options (changing the CIR and adjusting the connection price) and see if that “works” “economically.”
PS. I like the ISP Skytower POP, particularly the way that diagram illustrates the ease with which the International port can be the single locus of metering, if you must, along with the packet shaping, if you must.
Hamish.
LikeLike
Hamish, this model preserves CIR only through the operator’s network to its point of interconnection with other networks. When handed over to international, national peered, and national unpeered, the network becomes PIR/best effort. But that’s how it goes.
A model preserving current UBA options, delivered over UFB, would be profitable. It would also be putting lipstick on a pig.
LikeLike
Thank for clarifying that. My mistake.
As for the comparison, I’d like to see it, your example says there is no pig, lipstick or not.
And where is the 20:1 contention?
LikeLike
UBA contention ratios are the pig, running them over UFB would be the lipstick.
20:1 is the contention assumed at the handover between CFH and the ISP in the above case, the minimum handover speed required to provide 1000 subscribers with a 2.5mbps CIR.
LikeLike
I can shave $12k off your cost aswell, CIR on UFB is misleading, UFB GPON is mandated as a 24:1 split which gives each connection 100mbit/50mbit of uncontested bandwidth for each home. As for the NNI, Why would you oversubscribe that? $300 per 10gbit connection makes it attractive to run 100%.
You’re regional backhauling is VERY open, Transit from Northland to AKL is going to be much cheaper than transit from CHCH to AKL. You’re also ignoring providers who provide International Transit regionally negating the need for bigger intre-regional links. Take into account local IX traffic and local CDN nodes and the amount of bandwidth you need to buy changes alot.
Overall a nice attempt to simplify the complex world of ISP’s but it’s cheaper on some items and more costly on other’s 🙂
LikeLike