IXPS - THE "NO-BRAINER" SOLUTION TO KEEPING AFRICAN TRAFFIC LOCAL

Top Story

In most African countries if you send an e-mail across town it makes a long and circuitous journey to North America or Europe before going back to its intended recipient. It costs money in international connectivity charges and gives latency problems (a fractional but sometimes problem-causing delay). This inbuilt internet communications structure problem mirrors similar difficulties with Africa’s telecoms system: calls go north to get back to neighbouring countries. Local internet exchange points (IXPs) allow one country to route all (or most of) its internal internet traffic at a national level thus saving money and adding speed to the connection. In the not-too-distant future regional IXPs and perhaps even a continental exchange point might allow African countries to communicate across borders without going North. Perhaps with IP calling this would begin to also solve the parallel telecoms problem. But we’re getting ahead of the argument...Local IXPs are, as one African ICT policy advisor put it:"An absolute no-brainer." Russell Southwood and Brian Longwe explain why.

IXPs are the keystone of the entire internet economy: they interconnect different parts of the internet and they allow different ISPs to connect with each other, creating in effect a clearing house. Routing traffic the long way round is not an efficient way to use the network and thus the IXP mantra "keep local traffic local" developed.

Local IXPs offer both internet users and ISPs a number of distinct advantages:
- They improve quality by speeding up connection times: there is a 200-900 millisecond delay in each hop the message makes across the system, compared to 5-20 milliseconds locally.
- They save money because the calling costs are all at a local level.
- They create new revenue opportunities because for example local content providers can start creating things like locally hosted web sites, a range of e-services and streaming. The latter would simply not be feasible if an international connection had to be made.

The Kenyan IXP went online fully in 2002 with initially 4 ISPs but is now used by 10 of them. On an uncongested link, the latency is now 30-60 milliseconds. One rather conservative ISP decided that it would only require a 64k circuit to handle likely traffic and within two hours it was so packed that it got congested. Before it was established, international connectivity charges were nine times equivalent local costs. Within a very short period of time Telkom Kenya had slashed its international call rates in half.

There are now six IXPs in Africa: South Africa, Mozambique (opened May last year), Zimbabwe, Egypt (also May last year), Nigeria (Ibadan with only 2 ISPs) and the DRC (set up December last year). Uganda has one that is "semi-operational" but not all ISPs are using it. Tanzania keeps promising to open and will doubtless come on stream shortly. Other countries are holding preparatory discussions.

The major issue is one of trust. You need to be able to work with your competitors and in some countries this level of trust has not yet been established. As Brian Longwe told a recent workshop at the Southern African Internet Forum:"Getting any IX/peering arrangement off the ground is 10% technical work and 90% socio-political engineering." He also pointed out the importance of getting ("written") regulatory support. Setting up a local IXP is neither costly nor difficult.

Well over half of the connections to the North American or European "backbones" for the countries with under 5 ISPs are dominated by many of the usual suspects: France Telecom, Cable and Wireless, Teleglobe, Interpacket and Worldcom. Recent research claims that Africa is paying well over US$400 million for these connections. Whilst this figure is almost certainly open to challenge, it is easy to see that even a small share would lower the "operating costs" of the different countries of Africa, saving money and making them more competitive. Currently IXPs are carrying 20-25% of all traffic with the South African IXP carrying a much larger proportion.

If you buy the argument so far, let’s try the next two steps. It is technically not difficult to connect up the different local IXPs. For example, with these connections in place, Mozambique’s internet users could both e-mail and access the web in say South Africa without the traffic it generates leaving the continent, thus saving one of Africa’s poorer countries much needed hard currency. The barriers are largely political and regulatory. Indeed there’s nothing to stop there being an African internet exchange point that could exchange traffic between all countries on the continent.

The next much harder step is to imagine a world of African regulation that is "technology-agnostic", a world in which IP dialing or VOIP is possible. Calls could then be made using the internet that could be connected up using the regional IXPs or All-Africa internet exchange point. Just think of the scale of hard currency saved by not routing regional telephone calls out of the continent.

All of this is part of Africa’s future. The question is no longer will it happen but how fast can it be put in place? And the answers to that question will be in the hands of people and have much less to do with technology. Meanwhile local IXPs remain an absolute no-brainer...