Download to read offline
Uploaded byAFRINIC
Latency clustering AfPIF2017
This document analyzes latency patterns in Africa's internet through measurements from RIPE Atlas and Speedchecker probes. The key findings are: - Africa has higher average in-country latencies than other regions, around 78ms. - Countries cluster into Northern, Southern, Eastern and Western groups based on latency, though some exceptions exist like islands clustering with northern countries. - Inter-cluster latencies exceed 200ms, with the highest delays between the Western and other clusters. - Africa relies heavily on overseas transit providers, with over a third of routes first hopping outside Africa, primarily to Europe. This contributes to higher latencies. - Future work aims to better understand the role of undersea
More Related Content
Similar to Latency clustering AfPIF2017
Latency clustering AfPIF2017
- 1. Looking for Latency Clustersin Africa’s Internet AfPIF 2017 Agustín Formoso LACNIC Amreesh Phokeer AFRINIC Josiah Chavula AFRINIC Gareth Tyson Queen Mary University London Arjuna Sathiaseelan University of Cambridge 1
- 2. What We Know Circuitousroutes and high end-to-end latencies across the continent 2
- 3. Research Questions Which countriesare better inter-connected, which ones are not? – Key clusters of connectivity in the region – How are inter-country latencies impacted by topology and interconnection strategies? 3 AXIS Project
- 4. Measurement Platform Two platformsfor launching measurements ● 229 RIPE Atlas probes in Africa ○ 36 African countries ○ Hardware-based ● However: ○ mostly in university networks and ISPs? ○ about 50% of Atlas probes in Africa are in ZA ● 850 probes in Africa ○ 52 African countries ○ Software-based (Windows PCs) ○ Include edge networks / home users ● However: – No IPv6 – Unreliability (based on hosts’ availability) 4
- 5. ● Speedchecker Probes (www.speedchecker.xyz) – 850 software probes – 319 ASes – 52 countries ● 50% of the countries had at least 20% of their ASes probed. 5 Vantage points
- 6.
- 7. Data Collection • Pingfromeach probe: – select random African Speedtest server as target – launch 10 consecutive pings (one second apart) to their randomly chosen Speedtest server – return the minimum delay (RTT) observed at that time period • Traceroutefrom each probe: – launch a Traceroute to randomly selected Speedtest server – for each router hop • determine the ASN using the RIPE Routing Information Service • attach the geolocation using MaxMind GeoLite2-City 7
- 8. Data Collection 8 4 timesa day Pings and traceroute 3 months 42 200 RTT samples 31 500 traces captured 319 ASNs
- 9.
- 10. Country-level Latencies 10 Under 30ms ●Mauritius ● Reunion ● Cote D’Ivoire CI->CI <30ms Mean in-country Under 40ms ● South Africa ● Egypt ● Benin CD->CD ~300ms
- 11.
- 12. ● Northern ● Southern ●Eastern ● Western 12 Africa Latency Clusters
- 13. ● Northern ● Southern ●Eastern ● Western 13 Africa Latency Clusters EXPECTED
- 14. Unusual Cases inLatency Clusters ● Senegal, Liberia and Benin on the West coast, in Northern cluster ● Madagascar, Seychelles, islands of the Indian Ocean, clustered alongside countries in the North ● Somalia, on the East coast, is clustered with countries on the West coast. 14
- 15. Unusual Cases inLatency Clusters ● Senegal, Liberia and Benin on the West coast, in Northern cluster ● Madagascar, Seychelles, islands of the Indian Ocean, clustered alongside countries in the North ● Somalia, on the East coast, is clustered with countries on the West coast. 15
- 16. Unusual Cases inLatency Clusters ● Senegal, Liberia and Benin on the West coast, in Northern cluster ● Madagascar, Seychelles, islands of the Indian Ocean, clustered alongside countries in the North ● Somalia, on the East coast, is clustered with countries on the West coast. 16
- 17. Angola & Ethiopiaunclustered Ethiopia Upstream 100% Overseas Angola Upstream 50% Overseas* *16% Southern -> Europe 17 EU 354ms 144ms N.America
- 18. Inter-cluster Latencies inter-cluster medianRTTs Southern and Eastern ○ Lowest inter-cluster delay ~ 92ms ○ Lowest intra-cluster RTTs individually 18 Inter-cluster delays are mostly exceeding 200ms Western cluster has the highest intra-cluster RTTs
- 19.
- 20. Africa’s upstream providers ●37.8% of traceroute paths first hop outside of Africa ● 6.6% through SA and 4.5% through Mauritius (WIOCC) 20
- 21. Africa’s Top 10upstream providers 21
- 22. Africa’s Top 10upstream providers 22 France Telecom (Orange) accounts for 17% for French speaking countries in West Africa
- 23. Utilisation of upstreamproviders by cluster 23 50% have overseas hops (Europe) 14% with 4+ hops before Africa
- 24. Latency to Upstreamproviders 1. Same cluster 203ms 2. Other African Cluster 243ms 3. Overseas 268ms Northern North America: 71ms Southern North America: 227ms 24CDF of RTTs grouped by location of upstream providers RTT to the first AS hop
- 25. Take-aways 25 Below 40ms Benin Egypt Mauritius Ivory Coast Reunion SouthAfrica Cluster w/ high delays (Western) Due to higher usage of international transit providers Africa’s Interdomain communications highly reliant on inter-continental transit
- 26. Future work • Correlationwith Physical infrastructure, IXP presence in African countries and Peering relationships • Investigating delays between countries and popular web/content infrastructure • Linking the findings to regional Internet development strategies (e.g., deployment of regional IXPs) 26
- 27. THANK YOU! Questions? 27 Chavula J.,Phokeer A., and Formoso A. "Looking for latency clusters in Africa’s Internet." AFRICON, 2017. 13th Edition of IEEE AFRICON Conference in Africa. Cape Town, 18-20 September 2017. Preview: http://www.eecs.qmul.ac.uk/~tysong/files/africa-internet.pdf
- 28. Upcoming Internet Measurements Workshops Internet Measurement/RIPEAtlas Workshop NOV 27 - DEC 02, 2017 LAGOS, NIGERIA International Workshop on Internet Measurements Research in Africa DEC 11– 12, 2017 LAGOS, NIGERIA !!CALL FOR PAPERS UNTIL 17 SEPT 2017!! 28
- 29. • Central location(IXP, main ISP) • 10 Mbps dedicated • IPv4 (unfiltered) • IPv6 highly recommended (dual-stack) 29 Host an anchor !! Sponsored by AFRINIC !! [email protected]
- 30. References 1. Blondel, V.D., Guillaume, J.-L., Lambiotte, R., and Lefebvre, E. Fast unfolding of communities in large networks. Journal of Statistical Mechanics: Theory and Experiment, 10 (2008). 2. Chavula, J., Feamster, N., Bagula, A., and Suleman, H. Quantifying the Effects of Circuitous Routes on the Latency of Intra-Africa Internet Traffic: A Study of Research and Education Networks. 2015, pp. 64–73. 3. Fanou, R., Tyson, G., Francois, P., and Sathiaseelan, A. Pushing the frontier: Exploring the african web ecosystem. In World Wide Web Conference (WWW) (2016). 4. Formoso, A., and Casas, P. Looking for network latency clusters in the lac region. In Workshop on Fostering Latin American Research in Data Communication Networks (2016). 5. Fanou, R., Francois, P., and Aben, E. On the Diversity of Interdomain Routing in Africa. 2015, pp. 41–54. 6. Gilmore, J., Huysamen, N., and Krzesinski, A. Mapping the african internet. In Proceedings Southern African Telecommunication Networks and Applications Conference (SATNAC), Mauritius (2007). 7. Gupta, A., Calder, M., Feamster, N., Chetty, M., Calandro, E., and Katz-Bassett, E. Peering at the internet’s frontier: A first look at isp interconnectivity in Africa. Passive Active Measurement Conference (PAM) (2014), 204–213. 30