Africa’s Internet Exchange Points score well on lowering latency and local download speed but did not contribute to lowering of end-user costs, says new study
Those promoting Internet Exchange Points (IXPs) in Africa (and Balancing Act was one of them) made several broad arguments. They would enable cost savings as a larger proportion of traffic is exchanged using local rather than international bandwidth; They would improve access speeds for users and cut down delays in downloading through reducing latency. They would create revenue opportunities because they allow easier hosting of local domains and improved access speeds make certain types of applications possible. Thirteen years after the first IXP was launched in Johannesburg a research report has been published assessing their impact.
The purpose of this research entitled “Impact of IXPs – A review of the experiences of Ghana, Kenya and South Africa” commissioned by OSI was to look at the evidence for these three different kinds of impact outlined above. Crucially, whether the cost-savings IXPs may or may not have made helped local ISPs to pass on price changes to the end-user.
The proportion of local traffic going via a local IXP in Africa varies from 10-60%, depending on the scale of the market involved and the level of development in that market. Therefore three African countries were selected that fell at different points of development along this spectrum.
The researchers looked at a combination of the following as measures of impact: the price to end-users and the views of operators on price reductions; changes in access speeds; whether local content, hosting and applications had grown; and other external which may have had an effect. The core of the report is three case studies covering Ghana, Kenya and South Africa.
The conclusions drawn from the three country case studies were as follows:
* Even where the proportion of local traffic going via the IXP was high, as in the case of South Africa with 60% in 2008, operators claimed that savings made in bandwidth costs were insignificant. However, in this case, the cost savings did benefit Tier 2 ISPs who only had to pay to connect to one point (the IXP) rather than to several other providers. Where the Internet sector has been divided, as in Ghana, it has not been possible to gain the full cost advantages of a unified, single IXP.
* The original cost saving arguments for IXPs were predicated on there being a substantial difference between local, national and international wholesale charges. Since IXPs have been introduced, reductions in SAT3 bandwidth prices mean that in some cases national bandwidth costs may be the same or more than international bandwidth costs on a distance basis: for example, in Nigeria, it is cheaper to send traffic from Lagos to Sessimbra in Portugal than it is to send the equivalent amount of traffic from Lagos to Abuja. With the arrival of even cheaper fibre capacity in 2009 and 2010, this closing of prices will be a challenge for IXPs wanting to attract local traffic. But just as international prices have come down, so local and national prices will have to come into line with them. For IXPs to remain cost-effective, ISPs will need to press for lower national bandwidth charges.
• Since the retail cost of Internet subscription charges (and cyber-café access costs) in Ghana and Kenya has fallen since the introduction of IXPs, it could be argued that whatever cost reductions IXPs made for ISPs, they were passed on to subscribers or users. But since these reductions were also made by ISPs that were not IXP members as well, it is unlikely that they were as a result of savings from IXPs. In South Africa, dial-up charges have changed little over the past ten years and DSL charges have fallen since their introduction in 2003, well after the introduction of JINX. Again this makes it unlikely that any cost savings were connected with the existence of an IXP. Ghanaian and South African providers argue that cost savings were passed on to end users in the form of improved quality of service. But it is clear that cost savings from the IXP process were not passed on to end users except as part of the wider process of competition between operators.
• Whilst IXPs may have resulted in cost savings for the end user, the most significant factors have been: the ending of the international traffic monopoly (in Kenya), increased levels of competition and dramatic reductions in international bandwidth costs. For example in South Africa, international bandwidth as a percentage of total costs fell from 60% in 2003 to 45% in 2008. However, there is a direct link between IXP participants, their ISP associations and the bringing about of the factors listed above. For example in Ghana, GISPA has been instrumental in getting a special, low-cost deal on SAT3 bandwidth and in Kenya, TESPOK was at the forefront of the liberalisation process.
• Access speeds appear to have improved but it is difficult to separate the impact of improved national and international links from the speed advantage delivered by the IXP. Furthermore, as international bandwidth has come down in price, end users have had access to faster download speeds.
• Whilst the volume of traffic going through IXPs in Kenya and South Africa has increased dramatically, even these increased volumes have to a large extent been overshadowed by increases in international bandwidth. Indeed local traffic has fallen as a proportion of overall traffic in South Africa. There is no traffic measurement at the Ghana IXPs so it is not possible to say what has happened either in terms of the proportion of local traffic or the overall growth of traffic at the IXPs.
• Likewise the growth in local content and its use by end users has been eclipsed by a much wider interest in international content. As international fibre has become cheaper, it has been easier for operators to supply Internet users more cheaply. As a result, the number of people using the Internet has gone up. Although local content has grown in all three countries, the majority of use (particularly web mail access) remains international. For example, Facebook and You Tube are amongst the top ten sites accessed in African countries analysed by Alexa.com. However, this growth in Internet use has also benefited a small number of local sites. IXPs have supported the introduction of new local services and applications: for example in Kenya, the implementation of online tax reporting by the Kenya Revenue Authority and the availability of freeware hosted by the University of Nairobi.
The Impact of IXPs – A review of the experiences of Ghana, Kenya and South Africa was written by Charles Amega-Selorm, Muriuki Mureithi, Dobek Pater and Russell Southwood. If you would like a free copy of the report, please send an e-mail with IXP report as the header and we’ll send one to you.