No abstract

Despite extensive studies on the Internet topology, little is still known about the AS level topology of the African Internet, especially when it comes to its IXP substrate. The main reason for this is the lack of vantage points that are needed to obtain the proper information. From 2013 to 2016, we enhanced the RIPE Atlas measurement infrastructure in the region to shed light on both IPv4 and IPv6 topologies interconnecting local ISPs. We increased the number of vantage points in Africa by 278.3% and carried out measurements between them at random periods. To infer results that depict the behavior of ISPs in the region, we propose reproducible traceroute data analysis techniques suitable for the treatment of any set of similar measurements. We first reveal a large variety of ISP transit habits and their dependence on socio-economic factors. We then compare QoS within African countries, European countries, and the US to find that West African networks in particular need to promote investments in fiber networks and to implement traffic engineering techniques. Our results indicate the remaining dominance of ISPs based outside Africa for the provision of intra-continental paths, but also shed light on traffic localization efforts. We map, in our traceroute data, 62.2% of the IXPs in Africa and infer their respective peers. Finally, we highlight the launch of new IXPs and quantify their impacts on end-to-end connectivity. The study clearly demonstrates that to better assess interdomain routing in a continent, it is necessary to perform measurements from a diversified range of vantage points.

There is an increasing awareness amongst developing regions on the importance of localizing Internet traffic in the quest for fast, affordable, and available Internet access. In this paper, we focus on Africa, where 37 IXPs are currently interconnecting local ISPs, but mostly at the country level. An option to enrich connectivity on the continent and incentivize content providers to establish presence in the region is to interconnect ISPs present at isolated IXPs by creating a distributed IXP layout spanning the continent. The goal of this paper is to investigate whether such IXP interconnection would be possible, and if successful, to estimate the best-case benefits that could be realized in terms of traffic localization and performance. Our hope is that quantitatively demonstrating the benefits will provide incentives for ISPs to intensify their peering relationships in the region. However, it is challenging to estimate this best-case scenario, due to numerous economic, political, and geographical factors influencing the region. Towards this end, we begin with a thorough analysis of the environment in Africa. We then investigate a naive approach to IXP interconnection, which shows that a theoretically optimal solution would be infeasible in practice due to the prevailing socio-economic conditions in the region. We therefore provide an innovative, realistic four-step interconnection scheme to achieve the distributed IXP layout that considers and parameterizes external socio-economic factors using publicly available datasets. We demonstrate that our constrained solution doubles the percentage of continental intra-African paths, reduces their lengths, and drastically decreases the median of their RTTs as well as RTTs to ASes hosting the top 10 global and top 10 regional Alexa websites. Our approach highlights how, given real-world constraints, a solution requires careful considerations in order to be practically realizable.