-
Join the MyBroadband community
South Africa’s biggest forum. Discuss, discover, and connect with thousands of members.
-
Question of the Day: Which is the most hair-raising South African mountain pass you've driven?
Time for an open Internet exchange?
- Thread starter rpm
- Start date
The_Unbeliever
Honorary Master
It all comes down to expensive local bandwidth charges 
LazyLion
King of de Jungle
yes... it is time.
<moLe>
Expert Member
just do it peeps involved..its now or never
tjunyat
Senior Member
I could be entirely wrong, but I thought that CINX was going to scrap the equivalent line charges anyways if/when it came back online?
The issue of settlement free peering has always been a big one in SA and really should be addressed and implemented.
The issue of settlement free peering has always been a big one in SA and really should be addressed and implemented.
bekdik
Honorary Master
Is not Telkom the problem player?
Pilgrim
Wugger
Why don't the smaller players set up their own link (lets say 50 or 100 small players all link together) and then that one combined link connects to JINX or CINX?
That way they will carry a lot more weight.
That way they will carry a lot more weight.
Hi Bekdik
There are many issues associated with JINX, and Telkom’s rather high charges for local links do come into play (with Equivalent line charges). This charge varies from R 22,000.00 for a 2 Mbps to 8 Mbps link to R 375,000.00 for a 300 Mbps to 500 Mbps connection. These types of charges can obviously be prohibitive to a smaller ISP which wants to peer at JINX, but a new pricing structure is currently under discussion…
There are many issues associated with JINX, and Telkom’s rather high charges for local links do come into play (with Equivalent line charges). This charge varies from R 22,000.00 for a 2 Mbps to 8 Mbps link to R 375,000.00 for a 300 Mbps to 500 Mbps connection. These types of charges can obviously be prohibitive to a smaller ISP which wants to peer at JINX, but a new pricing structure is currently under discussion…
RoganDawes
Expert Member
Can't people see that BOTH parties lose by NOT peering?
Without local peering, both ISP 1 and ISP 2 have to pay for expensive international bandwidth.
Without local peering, both ISP 1 and ISP 2 have to pay for expensive international bandwidth.
On the topic of open IXPs, it seems some people are trying to revive The Hub in C.T., too. See the slides linked from here:
http://www.swimgeek.com/blog/2008/11/29/november-geekdinner-and-the-hub-project/
That only makes sense if both ISPs are roughly the same size, or if the bigger ISP is feeling generous towards the smaller ISP. If ISP 1 is much bigger than ISP 2, ISP 1 will tend to regard ISP 2 as a potential customer and won't be willing to give ISP 2 free access to their network. What tends to happen is that small ISPs peer with other small ISPs and buy transit from bigger ISPs, while big ISPs peer with other big ISPs.
In the rest of the world, peering is all about that sort of politics. In South Africa, the biggest problem is still getting from one's own network to the exchange because fibre connectivity is still unavailable or prohibitively expensive.
http://www.swimgeek.com/blog/2008/11/29/november-geekdinner-and-the-hub-project/
That only makes sense if both ISPs are roughly the same size, or if the bigger ISP is feeling generous towards the smaller ISP. If ISP 1 is much bigger than ISP 2, ISP 1 will tend to regard ISP 2 as a potential customer and won't be willing to give ISP 2 free access to their network. What tends to happen is that small ISPs peer with other small ISPs and buy transit from bigger ISPs, while big ISPs peer with other big ISPs.
In the rest of the world, peering is all about that sort of politics. In South Africa, the biggest problem is still getting from one's own network to the exchange because fibre connectivity is still unavailable or prohibitively expensive.
bekdik
Honorary Master
Wimbo
Well-Known Member
Some inaccuracies in the article
Nope, that's once off. You have the option to pay over 10 months if you wish.
Telkom's "rather high charges" are entirely the problem. Yes, we have a couple of alternatives (Neotel, MTN), but there isn't a significant cost saving. Hopefully, that will change early next year, when all VANS get IECNS licenses, and some of them start to build
real fibre networks.
Currently, the cost of getting your connection to JINX (or CINX) entirely dwarfs any other costs - ISPA membership, port charges, joining fees, etc. ISPA's highest membership fee at less than R7,000 is a whole heap less than R375,000 per month.
On Equivalent Line Charges (ELC), these were capped at R94,000 quite a while ago. You can see the traffic increases that have resulted here:
http://stats.jinx.net.za/showtotal.php?img=halfdecade
The peak for today is 278 Mbps.
Here's the irony: With the cost of an STM-1 (that's 155 Mbps) from JHB to London being only 1.6 times the cost of the same connection from JHB to CT, via Telkom, it's the local traffic that's expensive!
SA is unusual. Very unusual. The cost savings from peering are negligible. However, the improvement in latency and QoS are significant. So, for anyone operating their own network (and there aren't all that many in SA yet), it's worthwhile to peer.
The current situation at JINX is that any ISPA member can peer with any ISPA member who will peer with you. That's pretty open. ISPA doesn't force anyone to peer with you. ISPA doesn't force you to peer with anyone. ISPA encourages you to peer with everyone. You have to peer with at least 2 other parties to qualify.
ISPA is, however, working on making it even more open. Under discussion, for both JINX and CINX, is the ability for non-ISPA members to peer - again, with any one who's willing to peer with them.
The ELC is likely to disappear. The Joining Fee will probably go too. ISPA is considering moving to a Port Charge based model, with port speeds of 10 Mbps, 100 Mbps, 1 Gbps and 10 Gbps. Yes, ISPA has several takers already for 10 Gbps ports.
This will make life easier - and more rewarding - for anyone who wants to peer at JINX or CINX, whether they are an ISPA member or not.
JAWUG is welcome to apply to ISPA, either to join ISPA, or to peer at JINX & CINX, or both. ISPA currently has 158 members, and the number grows every month. By no means all of them are ISPs.
Nope, that's once off. You have the option to pay over 10 months if you wish.
Telkom's "rather high charges" are entirely the problem. Yes, we have a couple of alternatives (Neotel, MTN), but there isn't a significant cost saving. Hopefully, that will change early next year, when all VANS get IECNS licenses, and some of them start to build
real fibre networks.
Currently, the cost of getting your connection to JINX (or CINX) entirely dwarfs any other costs - ISPA membership, port charges, joining fees, etc. ISPA's highest membership fee at less than R7,000 is a whole heap less than R375,000 per month.
On Equivalent Line Charges (ELC), these were capped at R94,000 quite a while ago. You can see the traffic increases that have resulted here:
http://stats.jinx.net.za/showtotal.php?img=halfdecade
The peak for today is 278 Mbps.
Here's the irony: With the cost of an STM-1 (that's 155 Mbps) from JHB to London being only 1.6 times the cost of the same connection from JHB to CT, via Telkom, it's the local traffic that's expensive!
SA is unusual. Very unusual. The cost savings from peering are negligible. However, the improvement in latency and QoS are significant. So, for anyone operating their own network (and there aren't all that many in SA yet), it's worthwhile to peer.
The current situation at JINX is that any ISPA member can peer with any ISPA member who will peer with you. That's pretty open. ISPA doesn't force anyone to peer with you. ISPA doesn't force you to peer with anyone. ISPA encourages you to peer with everyone. You have to peer with at least 2 other parties to qualify.
ISPA is, however, working on making it even more open. Under discussion, for both JINX and CINX, is the ability for non-ISPA members to peer - again, with any one who's willing to peer with them.
The ELC is likely to disappear. The Joining Fee will probably go too. ISPA is considering moving to a Port Charge based model, with port speeds of 10 Mbps, 100 Mbps, 1 Gbps and 10 Gbps. Yes, ISPA has several takers already for 10 Gbps ports.
This will make life easier - and more rewarding - for anyone who wants to peer at JINX or CINX, whether they are an ISPA member or not.
JAWUG is welcome to apply to ISPA, either to join ISPA, or to peer at JINX & CINX, or both. ISPA currently has 158 members, and the number grows every month. By no means all of them are ISPs.
Drunkard #1
Expert Member
- Joined
- Aug 15, 2007
- Messages
- 3,668
- Reaction score
- 18
I think that the peering issues raised here are the core problem facing local Internet, not Telkom, or Ivy, or aliens. The ISPs are the consumer's worst enemy. In a previous article, Monochoice threatened to move its VOD service to international hosting. I assume that they looked into peering with the two biggest ISPs in the country, and decided that it was cheaper to host internationally. Whimbo mentions a link from Jo'burg to Cape Town. That's irrelevant, I'd bet that 90 % of South Africa's traffic originates or terminates within 10 km of Sandton. And now with more certainty regarding I-ECNS licences, intra Jo'burg links should be in the testing phase, ready to be lit up as soon as ICASA signs on the dotted line. A Motorola catalogue I looked at once, guaranteed 40 mbps wirelessly at 10 km for $ 20 000 per link. Illegal until now, but not illegal from mid January. JINX provides a good facility for local peering. It's now time for ICASA to enforce local peering, because the market has proven that it's not going to peer itself.
What exactly is a Line equivalency charge? If I've read the article correctly, it seems to be a pound you in the ass charge, imposed by JINX, on those ISPs lucky enough to avoid Telkom's pound you in the ass charge for a dedicated line. (Like JAWUG, who is legitimately transported over 802.11.) If JINX wants consumers to think of it as part of the solution and not part of the problem (which it is now), it needs to stop talking about eliminating charges, and start eliminating them.
What exactly is a Line equivalency charge? If I've read the article correctly, it seems to be a pound you in the ass charge, imposed by JINX, on those ISPs lucky enough to avoid Telkom's pound you in the ass charge for a dedicated line. (Like JAWUG, who is legitimately transported over 802.11.) If JINX wants consumers to think of it as part of the solution and not part of the problem (which it is now), it needs to stop talking about eliminating charges, and start eliminating them.
It is encouraging to hear these kinds of developments from ISPA. The 10 Mbps, 100 Mbps and 1 Gbps per models also sounds far more acceptable than the 128 Kbps, 512 Kbps model.
I know there have been calls for compulsory peering up to the lowest peering speed – ie 10 Mbps – which may create an ‘open’ system where smaller players have less trouble to peer with the big guns like IS, MTN-NS, Verizon, Neotel and hopefully in future Telkom and Vodacom.
One can however appreciate that running such a system is not easy and to keep everyone happy is a PR task of note – rather ISPA than me
I know there have been calls for compulsory peering up to the lowest peering speed – ie 10 Mbps – which may create an ‘open’ system where smaller players have less trouble to peer with the big guns like IS, MTN-NS, Verizon, Neotel and hopefully in future Telkom and Vodacom.
One can however appreciate that running such a system is not easy and to keep everyone happy is a PR task of note – rather ISPA than me
stoke
Honorary Master
I am loving this thread. This is where the 500mil thrown at Sentech should have gone.
Reality Check...
Guys,
It's all well and fine to blame carriers for high costs of leased lines and the likes, but let's face it, from where I'm standing (as someone who works with this stuff all day, every day) the truth of the matter is that ISPs simply aren't interested in peering with each other.
Nineteen (yes, that's 19) ISPs / WISPS have something in / on the building where our data centre is. This varies from a large data centre containing a couple of hundred servers, to just having a wireless access point on the roof, to having Diginet circuits into MUXs that are then split to various clients in the building. The bottom line is, they are in the building.
Our own IP is supplied by MTN Network Services. It's expensive, but you get what you pay for.
I'm not going to talk about email or http traffic in this post, but instead I'm going to use gaming as this is probably the biggest cause of frustration for end users. Once the user's pressed Send and Receive on his email client, he's done. It then becomes the ISPs problem.
A guy who's a client of WISP A and who's playing WoW or UT on one of our servers has to go through his WISP's network, on to IS, then to JINX, then to MTN NS, then to us.
A guy who's a DSL client of ISP B, who's playing the same game has his traffic routed via SAIX to MTN NS and then to us.
Some of the WISPs are starting to get clever and are buying simplex bandwidth from the likes of Sky Vision. Routing to these guys is hell. I've just done a quick trace route to one company I know who use Sky Vision and found the route to be MTN NS SA, MTN NS London, Tiscali London, Sky Vision New York, Sky Vision's UK earth station, WISP's earth station... 900ms later...and where's the client? Looking out of my window, I can see his grid antenna pointing to his WISP.
The first thing we need to remember is that peering is not only a mess in SA. Big traffic providers overseas also need to sort their issues out, but that doesn't solve the problem for Mr X who lives in the same security complex as me, uses WISP A and wants to play Wow on a server in our data centre, but has a longer ping time to the server than I do to the BBC servers.
Some time go, I decided to do something about peering, so I took a ProLiant server that was quite a beast, but that had been replaced with something bigger, fitted it with some Adaptec 4 port 100Mb LAN cards, loaded RRAS onto it and then contacted the 18 companies in the same building as us to find out if anyone would be interested in peering with us.
I have personally been in touch the technical guys from each of these organisations and asked them to peer with us. I even went so far as to draw up a presentation explaining the benefits derived from some simple RIP or OSPF routing. This isn't rocket science. We're talking about adding a couple of lines worth of commands to a router to tell it where it's neighbours are.
Fifteen companies told me to Foxtrot Oscar. Of these, four subsequently contacted me, trying to sell me IP.
Three companies were interested.
Of those, two are "still thinking" and one actively doing something about it. We should be peering with each other some time in January 2008. Their problem is that they need to run a very long length of fibre and need to drill a couple of holes in load bearing walls. The landlord has given permission for this after we provided them with an engineer's report, but the landlord wants a "professional" to drill the two 8mm holes.
Here's what I discovered on my travels.
First of all, the companies that are governed by management committees are a lost cause. As far as they are concerned "everyone who plugs into our network must pay".
Secondly there are the small WISPS who have a couple of DSLs to IS.
Some of these guys are keen on the concept, but when you explain to them that they'll have to do a bit of a redesign of their networks to get peering to work, they're just not interested. Many of these guys buy a DSL and connect the DSL to the wireless access point and that's the end of that. They don't even have internal backbones from one access point to another.
Some of these guys are actually running pretty fancy operations, their DSLs to IS are centralised in a data centre of sorts and they have backbones from one high site to the next. These guys are generally more technically oriented than the previous lot. Their excuse for not peering is usually that they don't have the money to add another router to their network to run a cable over to me (even if - as is the case with the one guy - that cable would be less than 20m in length and would entail drilling a hole in one wall). One WISP actually has quite an impressive setup that includes a Diginet to Verizon.
The last group are small organisations run by techheads like me. These are guys who are reselling DSLs and who also have wireless networks. These are the guys who can understand the benefit for their users and are the people who are willing to put their necks on the chopping block and do something about peering.
Guys,
It's all well and fine to blame carriers for high costs of leased lines and the likes, but let's face it, from where I'm standing (as someone who works with this stuff all day, every day) the truth of the matter is that ISPs simply aren't interested in peering with each other.
Nineteen (yes, that's 19) ISPs / WISPS have something in / on the building where our data centre is. This varies from a large data centre containing a couple of hundred servers, to just having a wireless access point on the roof, to having Diginet circuits into MUXs that are then split to various clients in the building. The bottom line is, they are in the building.
Our own IP is supplied by MTN Network Services. It's expensive, but you get what you pay for.
I'm not going to talk about email or http traffic in this post, but instead I'm going to use gaming as this is probably the biggest cause of frustration for end users. Once the user's pressed Send and Receive on his email client, he's done. It then becomes the ISPs problem.
A guy who's a client of WISP A and who's playing WoW or UT on one of our servers has to go through his WISP's network, on to IS, then to JINX, then to MTN NS, then to us.
A guy who's a DSL client of ISP B, who's playing the same game has his traffic routed via SAIX to MTN NS and then to us.
Some of the WISPs are starting to get clever and are buying simplex bandwidth from the likes of Sky Vision. Routing to these guys is hell. I've just done a quick trace route to one company I know who use Sky Vision and found the route to be MTN NS SA, MTN NS London, Tiscali London, Sky Vision New York, Sky Vision's UK earth station, WISP's earth station... 900ms later...and where's the client? Looking out of my window, I can see his grid antenna pointing to his WISP.
The first thing we need to remember is that peering is not only a mess in SA. Big traffic providers overseas also need to sort their issues out, but that doesn't solve the problem for Mr X who lives in the same security complex as me, uses WISP A and wants to play Wow on a server in our data centre, but has a longer ping time to the server than I do to the BBC servers.
Some time go, I decided to do something about peering, so I took a ProLiant server that was quite a beast, but that had been replaced with something bigger, fitted it with some Adaptec 4 port 100Mb LAN cards, loaded RRAS onto it and then contacted the 18 companies in the same building as us to find out if anyone would be interested in peering with us.
I have personally been in touch the technical guys from each of these organisations and asked them to peer with us. I even went so far as to draw up a presentation explaining the benefits derived from some simple RIP or OSPF routing. This isn't rocket science. We're talking about adding a couple of lines worth of commands to a router to tell it where it's neighbours are.
Fifteen companies told me to Foxtrot Oscar. Of these, four subsequently contacted me, trying to sell me IP.
Three companies were interested.
Of those, two are "still thinking" and one actively doing something about it. We should be peering with each other some time in January 2008. Their problem is that they need to run a very long length of fibre and need to drill a couple of holes in load bearing walls. The landlord has given permission for this after we provided them with an engineer's report, but the landlord wants a "professional" to drill the two 8mm holes.
Here's what I discovered on my travels.
First of all, the companies that are governed by management committees are a lost cause. As far as they are concerned "everyone who plugs into our network must pay".
Secondly there are the small WISPS who have a couple of DSLs to IS.
Some of these guys are keen on the concept, but when you explain to them that they'll have to do a bit of a redesign of their networks to get peering to work, they're just not interested. Many of these guys buy a DSL and connect the DSL to the wireless access point and that's the end of that. They don't even have internal backbones from one access point to another.
Some of these guys are actually running pretty fancy operations, their DSLs to IS are centralised in a data centre of sorts and they have backbones from one high site to the next. These guys are generally more technically oriented than the previous lot. Their excuse for not peering is usually that they don't have the money to add another router to their network to run a cable over to me (even if - as is the case with the one guy - that cable would be less than 20m in length and would entail drilling a hole in one wall). One WISP actually has quite an impressive setup that includes a Diginet to Verizon.
The last group are small organisations run by techheads like me. These are guys who are reselling DSLs and who also have wireless networks. These are the guys who can understand the benefit for their users and are the people who are willing to put their necks on the chopping block and do something about peering.
Ant_Brooks
Well-Known Member
Drunkard #1
Expert Member
- Joined
- Aug 15, 2007
- Messages
- 3,668
- Reaction score
- 18
Because capitalism isn't always able to manage an industry by itself. Markets aren't always free. Sometimes barriers to entry hinder competition. Sometimes the incumbents hinder competition. Anything that hinders competition is a threat to capitalism. When capitalism fails (like with monopolies and so on), regulators have to step in and impose conditions. Half of the time these conditions aren't counter to capitalism, but rather are required to ensure that markets are kept free and open.
So far, regulated capitalism is the best choice we have for advancing all stakeholders in society, rather than advancing a few, to the detriment of the rest. Corporates will always do what's best for their bottom line, and without regulation this abuse of the free market system will harm the populous. That is why regulation in this area is necessary, and beneficial to all stakeholders.