Economics of the UFB CIR

Forums › New Zealand Broadband › Economics of the UFB CIR

1 | 2 | 3

wjw

174 posts

Master Geek
+1 received by user: 5

#503645 9-Aug-2011 06:46

PenultimateHop: Very interesting statistics on the KAREN GGC consumption! Although I thought the first rule of the GGC is you don't talk about the GGC :-)

They shouldn't have made those stats public then

nickb800: I believe they have started to route general student traffic through KAREN, so that will be up to 15,000 students + staff watching youtube from Canterbury alone

You are correct, although it's the whole of the Karen network country wide.

Beccara

1473 posts

Uber Geek
+1 received by user: 517

ID Verified

#503696 9-Aug-2011 09:30

PenultimateHop:
Beccara: The traditional model of hauling everything back to Auckland for International handover and only have 3 main handover points of AKL/WGTN/CHCH is dead, We're going to need regional IX's to make this viable because my 50mbit/sec stream to the guy down the road going via AKL to get there simply isn't economical.

Things have to change, The old style of doing things just wont cut it and ISP's that don't adapt to it will be replaced somewhat by ones that do local caching/CDN/IX Peering/Targeted International bandwidth buys

It's not anywhere near that simplistic - yes, things will change but I'm very doubtful they'll change to that extent. The aggregate subscriber counts in New Zealand and the still relatively low per-subscriber bandwidth (even post-UFB) don't lend themselves to a highly distributed edge given the number of service providers. Even today you barely see ISPs with a highly distributed service edge, so getting to every ISP having "the big three" cities would be a huge improvement.

Moving gigabits around the country is actually relatively cheap vs. the capex and opex for deploying highly regionalized POPs.

wjw: Yea closest Limelight is Sydney. Had a long drawn out conversation with them trying to get them to come over here.... they require an aggregate of 1gbps of throughput from your AS to put a node on your network :-(

Also need a minimum of 100mbps aggregate throughput from your AS just to peer.

Precisely - this is why we won't see all of the "major CDNs" in NZ for a long time to come. From many CDN's perspective they already reach NZ for free through peering in Australia or Singapore, so why would they feel compelled to wear the cost to come closer?

Very interesting statistics on the KAREN GGC consumption! Although I thought the first rule of the GGC is you don't talk about the GGC :-)

I'm an optimist and maybe your right or maybe i'm right or maybe it'll end up somewhere in between us but under the UFB you have cheap 100/50mbit circuits on the last mile so I'm not sure why you are saying that post-ufb per subscriber bandwidth will be low. You are right ISP's currently don't have a highly distributed service edge but do you really think that model can survive the UFB?

CDN's may currently "get" NZ access via free peering in Australia and Singapore but if those CDN's want to improve or offer high bandwidth services then they will need to get closer. This is why we see under 10% of our traffic going to limelight and around 25% going to Google, Google's closer and plays ball more than limelight and this is reflected in our traffic. All it will take is one CDN willing to play ball with the UFB and the others should fall into line a bit more.

I know I'm dreaming when I say this but I do hope at some point down the road ISP's can work together a bit more, I'm confident that NZ as a whole push's 1gbit+ to limelight so if ISP's worked together we could get a node

Most problems are the result of previous solutions...

All comment's I make are my own personal opinion and do not in any way, shape or form reflect the views of current or former employers unless specifically stated

PenultimateHop

637 posts

Ultimate Geek
+1 received by user: 2

Trusted

#503700 9-Aug-2011 09:42

Beccara: I'm an optimist and maybe your right or maybe i'm right or maybe it'll end up somewhere in between us

Well I won't hedge either way but I have some experience in other markets.

Beccara: but under the UFB you have cheap 100/50mbit circuits on the last mile so I'm not sure why you are saying that post-ufb per subscriber bandwidth will be low.

Because they are low bandwidth circuits, and will be low bandwidth circuits when UFB actually penetrates a useful amount of the market. The average service is likely to be 30M or 50M and not 100M

Singapore has had 100Mbps services for years. So has Korea, Japan, Sweden and so on.

The average bandwidth per subscriber is still going to be relatively low in NZ. I've worked on networks that dimensioned for a 4.5Mbps average traffic flow per subscriber at the service edge - in 2006.

Beccara: You are right ISP's currently don't have a highly distributed service edge but do you really think that model can survive the UFB?

Yes I am quite sure it can. It's cheaper to backhaul GE (or even 10GE) to a larger POP than it is to deploy expensive equipment all over the place. The benefits of aggregation cannot be ignored, when you are talking about low density subscriber count that's highly distributed.

To crystallize the point a little more: a single modern BNG will terminate 128K subscribers. To make that equipment cost effective you need to amortize it across the largest number of subscribers you can - deploying 30 BNGs is not cost effective if (unless you happen to have 4M customers). You may choose (or need to) distribute where you have significantly high bandwidth per subscriber, but on average that's unlikely to occur when you consider that many ISPs in New Zealand do not have enough subscribers to fill even a single modern BNG to capacity.

Beccara: CDN's may currently "get" NZ access via free peering in Australia and Singapore but if those CDN's want to improve or offer high bandwidth services then they will need to get closer. This is why we see under 10% of our traffic going to limelight and around 25% going to Google, Google's closer and plays ball more than limelight and this is reflected in our traffic. All it will take is one CDN willing to play ball with the UFB and the others should fall into line a bit more.

I know I'm dreaming when I say this but I do hope at some point down the road ISP's can work together a bit more, I'm confident that NZ as a whole push's 1gbit+ to limelight so if ISP's worked together we could get a node

From the CDN's perspective they are getting everything they need: low latency ex-Australia and they can reach NZ via peering. The cost is on the ISP(s) in NZ and they will be the ones who get the end-user complaints, not the CDN.

ISPs in NZ have proven time and again for the last 15 years that they don't trust each other and won't work together. The market would be very different if they had and did.

Beccara

1473 posts

Uber Geek
+1 received by user: 517

ID Verified

#504041 9-Aug-2011 19:45

PenultimateHop:
Beccara: I'm an optimist and maybe your right or maybe i'm right or maybe it'll end up somewhere in between us

Well I won't hedge either way but I have some experience in other markets.
Beccara: but under the UFB you have cheap 100/50mbit circuits on the last mile so I'm not sure why you are saying that post-ufb per subscriber bandwidth will be low.

Because they are low bandwidth circuits, and will be low bandwidth circuits when UFB actually penetrates a useful amount of the market. The average service is likely to be 30M or 50M and not 100M

Singapore has had 100Mbps services for years. So has Korea, Japan, Sweden and so on.

The average bandwidth per subscriber is still going to be relatively low in NZ. I've worked on networks that dimensioned for a 4.5Mbps average traffic flow per subscriber at the service edge - in 2006.

Beccara: You are right ISP's currently don't have a highly distributed service edge but do you really think that model can survive the UFB?

Yes I am quite sure it can. It's cheaper to backhaul GE (or even 10GE) to a larger POP than it is to deploy expensive equipment all over the place. The benefits of aggregation cannot be ignored, when you are talking about low density subscriber count that's highly distributed.

To crystallize the point a little more: a single modern BNG will terminate 128K subscribers. To make that equipment cost effective you need to amortize it across the largest number of subscribers you can - deploying 30 BNGs is not cost effective if (unless you happen to have 4M customers). You may choose (or need to) distribute where you have significantly high bandwidth per subscriber, but on average that's unlikely to occur when you consider that many ISPs in New Zealand do not have enough subscribers to fill even a single modern BNG to capacity.

Beccara: CDN's may currently "get" NZ access via free peering in Australia and Singapore but if those CDN's want to improve or offer high bandwidth services then they will need to get closer. This is why we see under 10% of our traffic going to limelight and around 25% going to Google, Google's closer and plays ball more than limelight and this is reflected in our traffic. All it will take is one CDN willing to play ball with the UFB and the others should fall into line a bit more.

I know I'm dreaming when I say this but I do hope at some point down the road ISP's can work together a bit more, I'm confident that NZ as a whole push's 1gbit+ to limelight so if ISP's worked together we could get a node

From the CDN's perspective they are getting everything they need: low latency ex-Australia and they can reach NZ via peering. The cost is on the ISP(s) in NZ and they will be the ones who get the end-user complaints, not the CDN.

ISPs in NZ have proven time and again for the last 15 years that they don't trust each other and won't work together. The market would be very different if they had and did.

For a start UFB circuit tails are 30/10 or 100/50, There is no 50mbit circuit (Unless the RSP build a custom product out of EIR blocks) so speeds above 30mbit and below 100mbit are constraints places by the ISP AND the 30/10mbit circuits pricing increases over time while the 100/50mbit circuit price drops in order to shift consumers on to the faster tails.

We're going to have to agree to disagree about distributed edge access, If you've seen the pricing on hauling data around NZ recently you'll know that keeping just 1gbit of traffic in the same region can save you $100k+ per year and there are already a number of ISP's that operate at both APE and WIX I think you're vastly overestimating the cost of BNG/BRAS equipment deployment.

CDN's may see it right now as getting everything they need as currently the choke point is the last mile access but someone will ask the question either internally or it will come from a client as to why clients with 100mbit to the ISP are not getting anywhere near that. All it takes is one to lead the rest of the pack for the most part (Limelight excluded)

iSKY is an example for us, They spit out the content in Whangarei for us in order to provide the best possible speeds to the end user and to give ISP's less reason to not zero-rate it. There is also another CDN that's paid access but that access is cheaper than hauling that data down to Auckland and we see that as data clients want.

So whilst at the start ISP's may just haul it all back to AKL there will come a point where they have to look at better ways of managing the volumes of traffic they are handling because consumers will be wanting faster net for less money and they will move providers to get this albeit somewhat slowly but they will

PenultimateHop

637 posts

Ultimate Geek
+1 received by user: 2

Trusted

#504049 9-Aug-2011 19:58

Beccara: For a start UFB circuit tails are 30/10 or 100/50, There is no 50mbit circuit (Unless the RSP build a custom product out of EIR blocks) so speeds above 30mbit and below 100mbit are constraints places by the ISP AND the 30/10mbit circuits pricing increases over time while the 100/50mbit circuit price drops in order to shift consumers on to the faster tails.

Right and I would expect the 30M to be the most typical tail initially, and we are talking significant number of years before substantial amounts of NZ will have access to that. 50Mbps was a brain-fart but it wouldn't surprise me to see a 50Mbps service offered (even done by consuming a 100Mbps tail) at a retail level.

Beccara: We're going to have to agree to disagree about distributed edge access, If you've seen the pricing on hauling data around NZ recently you'll know that keeping just 1gbit of traffic in the same region can save you $100k+ per year and there are already a number of ISP's that operate at both APE and WIX I think you're vastly overestimating the cost of BNG/BRAS equipment deployment.

Well, I've done significant modelling of networks in New Zealand and other parts of the world and I'd be very surprised if a ~30 POP service edge ever made sense in New Zealand. You DO NOT avoid the cost of backhaul with a highly distributed edge, you may slightly reduce it but the increased capital and other operational costs will significantly diminish or completely negate the backhaul savings.

To be clear: I never said ISPs would continue to keep everything in Auckland and I do anticipate seeing a distribution to at least the three biggest cities, but I am cynical about seeing much further distribution except for the very largest of ISPs (or ISPs that have found a significant non-retail cornerstone customer that justifies the development of significant POP assets in a place like Nelson). Remember that the cost of backhaul is most likely to significantly decrease as UFB comes online due to both demand and availability, as well as a little more competition.

I doubt I'm overestimating the cost of BNGs - installing them is what I do for a living.

Disagree all you like. :-)

Beccara: CDN's may see it right now as getting everything they need as currently the choke point is the last mile access but someone will ask the question either internally or it will come from a client as to why clients with 100mbit to the ISP are not getting anywhere near that. All it takes is one to lead the rest of the pack for the most part (Limelight excluded)

I'll disagree that the last mile access is the only choke point at present, and having worked for a CDN in the past I'm not sure that international operators are so likely to jump into the NZ marketplace, particularly with an interesting proliferation of NZ-based CDNs likely to start in the near future. If international CDN operators do deploy it will be to one or two strategic IXPs, rather than deploying 50 zillion locations to every ISP with a regional POP.

Your iSKY comments are interesting. I'm sure we'll see more of that for niche CDNs, although at a certain point they become difficult to justify on an ROI perspective.

Beccara: So whilst at the start ISP's may just haul it all back to AKL there will come a point where they have to look at better ways of managing the volumes of traffic they are handling because consumers will be wanting faster net for less money and they will move providers to get this albeit somewhat slowly but they will

I'm happy to be proven wrong. But I don't see it. Global experience does not gel with your proposal - see the anguish in Australia over the NBNCo POIs as one amusing example close to home; elsewhere in the world it is typical to see IP POPs located in the most strategic place for maximum aggregation at minimum cost.

Beccara

1473 posts

Uber Geek
+1 received by user: 517

ID Verified

#504058 9-Aug-2011 20:24

Reading over my post's It's clear I haven't stated this either but I'm not saying we will see a IX at every UFB POI but I do expect we will see more than AKL/WGTN/CHCH.

The KAREN map's with GCC are interesting, Here you have a network of a few people with high speed access pushing 600mbit peak to a cache so not a CDN, If this were contained in one or two UFB zones why would you not deploy your edge to this so you don't have to haul upto 600mbit/sec back to your GCC node in AKL? And thats just Google cache, Chuck in a Akamai node and you may very well see this up over 1gbit/sec

Cost of backhauling may come down a little but we already have competition in this area, There are 3(? or 4) providers between AKL and WGTN, I know up in Northland we have 3 options for hauling down to AKL. It may come down enough that in the short/medium term the deployment cost's of the edge may be more than the cost of transit but there will always be a point at which it's cheaper to deploy the Edge to the location than to keep buying a bigger pipe

Remeber ISP's offering nationwide UFB services are already forced to have 30+ POI's with some level of gear and some level of operational cost's.

1 | 2 | 3