BTTB,
On the unshaped ADSL account:
In my humble opinion, absolutely nothing. Since real latency doesn't seem to be effected by being shaped or not shaped, I can't see how it would improve your experience.
For forex traders, it might offer a small improvement in terms of the actual bandwidth they can force through the system. This means, large graphs might be able to download to their systems faster. In terms of latency, I don't think it will make any difference at all.
Those who will gain the most from the unshaped account are not gamers or forex traders, as Telkom's misleading advertising would lead us to believe, but people who frequently use P2P file sharing networks.
Since the removal of traffic shaping would allow P2P traffic to gain a high network priority, throughput should improve dramatically for P2P users. Of course, once you reach 4GB of traffic volume, you can kiss your P2P downloads goodbey for the month.
On a whole, I would recommend that everyone stay away from this new unshaped service. For the huge price difference it is definitely not worth it.
On the topic of PPPoE server locations:
First off, let me apologise to everyone in Cape Town. The PoP is in fact in Barrack Street, not Barrat Street. I don't know where I came up with the incorrect name, though I vaguely remember reading it off a SAIX code allocation list a very long time ago.
Now, as for moving all the PPPoE servers to Barrack Street. This might not be a bad idea at all. It is not the most elegant solution, as it would decrease international latency but increase local latency, but it would definitely be a good first aid measure to improve the international latency for ADSL users.
I would go one step further though. Simply establish a PoP in Melkbosstrand and host the PPPoE servers there, literally on the SAT3 end point.
The best way would still be to split the network though. While working this afternoon on a classic car engine I am restoring, I was thinking about my previous post and it did occur to me that there is a practical way to allocate VCI numbers to each area. The VCI could be the same as the area code of the physical PoP where the PPPoE server is hosted.
For instance, for the Western Cape, with the PoP in the 021 area code, VCI could be set to 21. For the Free State, with the Bloemfontien PoP in the 051 area code, the VCI could be set to 51. We could also have 11 and 12 for Gauteng, 31 for Durban and so forth.
This would still require that SAIX set up PPPoE servers at each PoP and that their helpdesk staff be trained to help users set up the correct VPI and VCI, but it would at the very least not be too confusing.
The reason I say all this is the way local and international traffic is generally routed in South Africa. In the case of ADSL users, all international traffic is routed via SAT3, from Cape Town.
Thus, locating the PPPoE servers in Cape Town would seem to be a very good idea. However, this will cause local traffic to suffer for users in the rest of the country, inversely to the way international traffic from Cape Town suffers now.
The ISPA (Internet Service Provider Association of South Africa), the group that encourages co-operation between South African ISPs, has its main peering point in Johannesburg. The JINX (Johannesburg InterNetworking eXchange) provides a crucial local routing stop-off for all major ISPs.
Traffic from one local ISP to another must pass between the networks somehow. Before JINX and CINX (the Cape Town based emergency backup to JINX) were established in 1996, any traffic between different South African ISPs would first need to travel to their upstream providers overseas, then over the overseas internet to the second upstream provider and return via international links to the second local ISP.
JINX solved the problem by allowing local traffic to pass between different IP networks without the need to leave our shores.
If all PPPoE servers were to be moved to Cape Town, local latency for Gauteng users would be adverseley effected.
For instance, when connecting to a web site hosted in Johannesburg, a user from Pretoria would first have his traffic routed down to Cape Town on the ethernet over ATM layer, through the PPPoE server, then back to Johannesburg on the IP layer.
Also, anyone browsing local sites on hosted on networks other than SAIX would really feel the effects of such a move. If all PPPoE servers were in Cape Town, someone in Gauteng browsing a site hosted on the IS network, from an ADSL connection to the SAIX network, would have their traffic routed as follows: From Gauteng to Cape Town, through the PPPoE server, back to Gauteng, through JINX, to IS.
Obviously, we would be sacrificing good local latency for improved international latency.
Pondering this all has lead me to another interesting hypothesis which I will share with you now. If Telkom were lazy about setting up their ADSL PPPoE servers, they would naturally have chosen to host them all in a single location.
It should be fairly obvious to all by now that the ADSL network is optimised to carry local traffic and that international traffic is second on the list of priorities. If Telkom decided to be lazy about the placing of PPPoE servers, Gauteng would be the natural choice for the location of these servers, as placing servers close to JINX would mean the best possible local latency under the conditions.
If all PPPoE servers were to be moved to Cape Town without severely impacting local latency, the ISPA would also have to change its primary peering point from JINX to CINX. This is very unlikely though, since CINX is a small back-up peering point only used by the largest ISPs. Lines running into CINX are generally extremely small, rarely more than 128kbps, as they are only used for back-up perposes. Lines running into JINX range from 2MBps to multiple 34MBps connections.
Moving the primary peering point would thus mean that all ISPs will have to reshape their network topology and install large capacity lines in Cape Town. It will also be a disaster for most web hosting providers, as they have all gone to great trouble to invest in infrastructure in Johannesburg, to be close to JINX.
Giving each area its own PPPoE server and splitting the network would solve both these problems. Since all traffic will be able to take the shortest possible route to both critical routing points, the SAT3 endpoint and JINX. Local and international latency for all ADSL users would be greatly improved.
Perhaps, we might write up a letter to Telkom suggesting such a change, I doubt we would even be considered though, as the scale of the change needed to split the network now will be well beyond the investment in man hours that Telkom would be prepared to make. Not to mention the disruption the change will cause to those users who either ignore the request that they change their settings, or simply do not know how.
Willie Viljoen
Web Developer
Adaptive Web Development
*
]
] 