Time for an open Internet exchange?

[stoke]

@ portcullis : OMW ... that is absolutely and truly pathetic to say the least. And, when you get right down to the bottom of the mess, the only ones winning here are the bulk resellers overseas. Stupid stupid stupid.

LOCAL TRAFFIC HAS TO ROUTE LOCALLY, I do not send a snail mail to my next door neighbor via new york. I get up and go and speak to her.

Perhaps the consortium of local-accounting-routers should be setup that buys up all the routing equipment and re configures them all to peer, and then gets funded using basic "I sent 1mb to you this month, so I owe you 1mb, but you grabbed 3mb from me, so you owe me 2mb.". The only catch with this approach is that the heavy data senders will be scoring while the light data senders will be losing.

Is this the crux of the problem? ISP's are scared that another interlinked ISP will consume all their available throughput resources, without compensation?

Well, then they are perfectly entitled to ask for compensation, and at the same time, they're also encouraged to host data-intensive services that they will then get paid for as other use these services.

One danger here is that a trend towards data heavy services may raise it's ugly head.

I think I see where the origination of a "private internet" where you pay for the content you access came from.

Perhaps it's time the industry faced the facts and started handling large traffic (e.g. downloading movies) as a separate pay service network.

 

[Ant_Brooks]

I find myself agreeing with everything you wrote there, Drunkard #1. But the argument that markets should (in general) be subject to some level of regulation isn't an answer to my specific question.

Why should it be compulsory for the "big guns" to peer with the smaller players?

Is there any evidence of market failure in the Internet sector when it comes to interconnection? Do smaller ISPs have evidence of anti-competitive behavior from the "big guns"? Do you perhaps think that ICASA's pending interconnection regulations do not already sufficiently address interconnection between ECNS and ECS licensees?

 

[portcullis]

Why should it be compulsory for the "big guns" to peer with smaller players?

Attitudes like yours are hindering the growth of the internet in South Africa.

 

[portcullis]

Is this the crux of the problem? ISP's are scared that another interlinked ISP will consume all their available throughput resources, without compensation?

I suspect that many of the players don't understand the concept of peering and confuse peering with pooling.

One of the things that surfaced time and again when I was talking to ISPs about peering was "what guarantees do I have that you won't steal my international bandwidth?".

They don't want to understand that all we are interested in is getting to their servers, not their international links.

 

[portcullis]

Do you perhaps think that ICASA's pending interconnection regulations do not already sufficiently address interconnection between ECNS and ECS licensees?

How many VOIP service providers are interconnected?

Aren't there regulations in place governing these interconnects, that are simply being ignored?

 

[Ant_Brooks]

Attitudes like yours are hindering the growth of the internet in South Africa.

Excuse me? What attitude? I didn't express a point of view, I asked a perfectly valid question.

 

[Ant_Brooks]

How many VOIP service providers are interconnected?

That's at least two questions bundled into one. Do you mean:
- How many ISPs that provide VoIP services are interconnected with each other?
- How many ISPs that provide VoIP services are interconnected with traditional voice operators (such as Telkom and the mobile operators)?

Aren't there regulations in place governing these interconnects, that are simply being ignored?

Not really. There are interconnection regulations (possibly still in draft format; I don't recall offhand if those have been finalised or not) which are applicable to ECS and ECNS licensees. But ICASA has not yet finished converting VANS licences into EC(N)S licences, which means the interconnection regulations remain in limbo until they've done so.

 

[portcullis]

Excuse me? What attitude? I didn't express a point of view, I asked a perfectly valid question.

Because, until such time as the "big guns" acknowledge the smaller players and allow them their little place in the sun (something the "big guns" appear to try their utmost to avoid) nothing will change.

 

[portcullis]

That's at least two questions bundled into one. Do you mean:
- How many ISPs that provide VoIP services are interconnected with each other?
- How many ISPs that provide VoIP services are interconnected with traditional voice operators (such as Telkom and the mobile operators)?

I'm specifically referring to voip providers.

Case in point. As of a minute ago, our Telfree 087 line could not dial our Switch Telecom 087 line.

 

[Ant_Brooks]

Because, until such time as the "big guns" acknowledge the smaller players and allow them their little place in the sun (something the "big guns" appear to try their utmost to avoid) nothing will change.

Sorry portcullis, but "they should allow smaller players their place in the sun" is hardly a well-justified and cogent argument for forcing large ISPs to interconnect with small ISPs!

Let me be clear: I agree that more interconnection between ISPs -- in general -- is a good idea. In most cases (but not all), I think it makes sense for two ISPs to interconnect with each other irrespective of their relative sizes.

But I'm not willing to leap from "I think it is usually a good idea for two ISPs to interconnect" to "all ISPs should be forced to interconnect with each other". In general, I'm a fan of light-touch regulation. Far too many laws have noble intentions, but end up being counter-productive. So if there need to be regulations forcing ISPs to do something (such as interconnect with other ISPs) then there had better be damn good reasons and justification for those regulations.

And there might well be some good reasons, but so far, I don't think they have come across very clearly in this thread.

 

[Drunkard #1]

WRT your earlier query, I'm not involved in the ISP field, so it is quite onerous to keep up to date with draft regulations and so on. My main argument is based in the pricing of these interconnects. Providing solid evidence for this observation is impossible. Most (all?) of these interconnect agreements are private business deals, and thus secret and probably covered by non disclosure agreements. However, my interpretation of Monochoice's recent comments, is that hosting internationally is cheaper than peering locally. And I don't think that they're so short sighted that they'd ignore what's going to happen in mid January.

As for forcing everyone to peer with everyone else; the biggest ISP will peer with everyone and their dog if it was in their financial interests, so the question isn't whether they will peer, but at what price. This is where regulation is needed, because as with cellular interconnect fees, Internet peering fees would lead to anti-competitive practices and to large ISPs abusing their monopolies. Even more so than with cellular, because there are only three players in the cellular industry, but 600 in the ISP industry. These charges need to be cost based and not earnings based. I E : What does it cost an ISP to serve 1 Gbps of data (to a point in the same building), plus a bit of profit, rather than what can the market tolerate. These issues won't be resolved by market forces (as we've seen with VANS and cellular) and thus need to be regulated.

 

[Ant_Brooks]

As for forcing everyone to peer with everyone else; the biggest ISP will peer with everyone and their dog if it was in their financial interests, so the question isn't whether they will peer, but at what price. This is where regulation is needed, because as with cellular interconnect fees, Internet peering fees would lead to anti-competitive practices and to large ISPs abusing their monopolies. Even more so than with cellular, because there are only three players in the cellular industry, but 600 in the ISP industry. These charges need to be cost based and not earnings based. I E : What does it cost an ISP to serve 1 Gbps of data (to a point in the same building), plus a bit of profit, rather than what can the market tolerate. These issues won't be resolved by market forces (as we've seen with VANS and cellular) and thus need to be regulated.

What you seem to be saying here is that large players should not be able to charge extortionate fees to smaller players for the privilege of interconnection, and that such agreements should be cost-based and not unreasonably withheld. I totally agree with that, and it is also (more-or-less) what ICASA's interconnect regulations mandate, albeit without the tonne of associated bureaucracy that the regs contain.

This is also a much more reasonable position that "all large ISPs should be forced to interconnect with all small ISPs at no charge".

 

[Drunkard #1]

What you seem to be saying here is that large players should not be able to charge extortionate fees to smaller players for the privilege of interconnection, and that such agreements should be cost-based and not unreasonably withheld. I totally agree with that, and it is also (more-or-less) what ICASA's interconnect regulations mandate, albeit without the tonne of associated bureaucracy that the regs contain.

This is also a much more reasonable position that "all large ISPs should be forced to interconnect with all small ISPs at no charge".

I don't think that "for no charge" is a bad model. ISPs generally have two sources of income: hosting and access. Surely the income they receive from hosting content should cover the costs of serving that content to a point in the same building. They have not incurred any of the expenses associated with providing access (for the end user) to that content, and therefore aren't entitled to any of the revenue. I'm sure self generated content could be worked into the model somewhere.

This isn't an about face either - A cost based model will predict prices in the order of fractions of a cent per GB. Which is why I endorse the more easily billed pipe size model over the volume model (charge per Gbps, whether its used or not, don't charge per Gb) And even then the charge shouldn't be more than a few grand a month.

Also, 10 Mbps is a step in the wrong direction. That's 2.5 ADSL lines at full speed (or 0.5 UK standard ADSL lines). Please. 1 Gbps and up. We're trying to progress here, not provide the lowest possible speed that the market will tolerate.

 

[Ant_Brooks]

I don't think that "for no charge" is a bad model. ISPs generally have two sources of income: hosting and access. Surely the income they receive from hosting content should cover the costs of serving that content to a point in the same building. They have not incurred any of the expenses associated with providing access (for the end user) to that content, and therefore aren't entitled to any of the revenue. I'm sure self generated content could be worked into the model somewhere.

I don't quite follow this. Let's say that there is a large ISP named "Largeco" and a much smaller ISP named "Smallco". Assume for the purposes of this example that both of them happen to be based in the same building. Smallco happens to be a wireless ISP focussing on the niche market of online gamers in a particular city (Cape Town, say). Largeco services more of a corporate market, and has a national network spanning all nine provinces. But they also service a fair number of gamers in JHB (say). Now, let's assume a customer of Smallco in CT and a customer of Largeco in JHB are busy with some fairly bandwidth-intensive gaming. Assuming that Largeco is forced to interconnect with Smallco at no charge, how does the cost of carrying packets between those two customers break down?

Well, Smallco is responsible for moving the packets from their customer to the point of interconnection with Largeco in Cape Town. Largeco has to transport those packets from the point of interconnection on its network in CT, across its national backbone to JHB, and then to the customer in JHB. The cost of long-distance fixed lines in South Africa is currently extremely high. In many cases it is more expensive to move a packet from JHB to CT than it is to move a packet from JHB to London. And Largeco is paying the entire cost of moving the packets cross country between the two customers. Smallco is contributing nothing to those costs.

Why should Largeco not be able to charge Smallco a reasonable cost-based interconnection fee to cover the cost of carrying Smallco's packets across Largeco's national network? Both ISPs are providing the same service to their customers. Why should one of them be forced to pay an unfair share of the cost of connecting those two customers?

 

[Drunkard #1]

There are various facets to this question.

1. If Largeco's JHB based customer was accessing a website hosted in their CPT offices, Largeco still paying for that JHB-CPT link, and is still out of pocket. As far as their customer is concerned, everything in SA is "local". He doesn't see "national" and "local" as two separate types of access (in the same ways as he differentiates "local" from "blended"). I would guess that due to the minor impact of local peering, only content hosted on his Largeco's own servers is "local", while content hosted on OtherISP's servers in the next room still has to cross a national network. I E, because only a very small percentage of Largeco's traffic is truly local, they don't differentiate between "local" and "national".

2. If Smallco peers with Largeco in CPT, then the peering agreement should only cover locally (CPT) hosted content. Smallco has two options here, either peer with Largeco in JHB (via it's own dedicated link to it's JHB peering point) or to pay Largeco for access to it's national network.

This charge should be market related, but CINX should have a huge pipe to JINX available anyone to use, at low cost. This is why peering points are important in the network. They allow everyone to install just one pipe to the peering point rather than several pipes to each peer, and they should provide major links at virtually no cost due to their bulk buying power.

3. How much national traffic has a different city for it's points of origin and termination. I would guess that 90 % of South African originates or terminates in JHB, and 50 % of South African traffic originates AND terminates in JHB. This is just a guess, but if it's true, just one peering point (JINX) should reduce ISP costs (and therefore charges) significantly. JHB - DBN links would follow, then JHB - CPT, followed by the rest.

 

[portcullis]

The cost of long-distance fixed lines in South Africa is currently extremely high. In many cases it is more expensive to move a packet from JHB to CT than it is to move a packet from JHB to London.

Do MTN, Vodacom and Neotel pay ICASA directly for having their own circuits between cities or are these circuits covered by their license fees?

Apart from the capex of building towers, getting Uncle Andrew to come out with his 6m Microwave dishes and his multi GHz transceivers and the ongoing cost of having people to monitor the links, what recurring costs are there?

I agree that even a 2Mb Diginet from Cape Town to Johannesburg would be overpriced, but what about these guys who have built their own networks? Be that microwave or fibre optic.