East Africa has been hit hard by these breaks, but in fact, it's easy to lose sight of how far Internet connectivity in this region has come in a few short years. It's stunning to see how much of the Internet infrastructure in Kenya and Uganda is actually still up and running, although congested. East African ISPs have learned a tremendous amount in a short time about designing their networks for resilience and diversity, and that's evident in the data pertaining to these cable breaks.
The graphic on the right illustrates the impact of the most recent TEAMS cable cut. Our data peg the time of the cut at 09:13 UTC on February 25, 2012, as providers throughout the region see sudden, massive shifts in their international Internet connectivity. Kenya normally has about 550 routed networks, and by this measure, over 60% of the country's Internet became unreachable at this point. Uganda normally has about 180 routed networks, and experienced a similar degree of impairment. Rwanda and Tanzania do not appear to have been affected by this cable cut, which is not surprising, since TEAMS is a point-to-point cable connecting Kenya to the United Arab Emirates (see Telegeography's online Submarine Cable Map).
The TEAMS cable was severed 5 km off the East African coast last Saturday, reportedly the result of an anchor drop. TEAMS' General Manager Joel Tanui has stated that it may take up to three weeks to repair the damage.
What would you expect to see at the moment of the cut? When a submarine cable is broken, international Internet providers that sell their services over that physical connection become inaccessible to local providers. If a local Internet provider was clever enough to buy transit from multiple service providers (a practice known as "multihoming"), and if those service providers are available on different cables, then Internet service fails over to those other cables.
Some Internet providers in Kenya and Uganda did exactly that: anticipating the possibility of failure on TEAMS, they bought transit on alternative cable systems (SEACOM or EASSy) as well.
Some providers are tied to particular cables, usually because they are part of a consortium that owns and operates the cable. So when that cable breaks, its associated service providers lose money, as traffic shifts to other alternatives. In this case, for example, Kenyan and Ugandan providers lost access to UAE-based Etisalat (AS8966), which is a major owner of the TEAMS cable. Many multihomed providers elected to send more of their traffic to international markets via Tata (AS6453) which participates in the SEACOM cable system. We call this kind of effect a "transit shift" — literally, watching local providers change their connectivity preferences in order to keep their customers connected.
In the graphics below, we can clearly see the Etisalat bands disappear as the cable is broken. Each dot represents a cluster of traceroutes which shared the same overall latency at around the same time. The colors represent the provider hand-off that was traversed. For example "8966 33771" represents traceroute latencies which crossed Etisalat (AS8966) to get to Safaricom (AS33771).
African providers who anticipated cable damage and engineered for resiliency actually benefited from the break. Yesterday CIO East Africa reported that, "Kenya Data Networks, also a 10 per cent shareholder of TEAMS, says that its customers were barely affected as it equally routes its traffic through the three cables in Kenya - TEAMS, EASSy and SEACOM." Kenya Data Networks (AS33770) not only survived the cable cut, they actually gained traffic, as other providers rerouted their connections through them. On the right, we can see that the number of successful traceroute measurements per hour across AS33770 increases after the cable cut.
Kenyan incumbent Kenyan Post & Telecom (AS12455) survived the break by shifting away from Etisalat on TEAMS to SEACOM.
The same effects are visible with Kenyan Post & Telecom, who actually gained traffic from other local providers, rerouting their traffic to reach the outside world.
In Uganda, Orange Uganda (AS36991) was one of those providers who maintained its connection to the Internet through the Kenyan Post & Telecom (AS12455), even as they lost their connection to parent company Orange.
Elsewhere in Uganda, MTN Uganda's primary ASN, 16637, dropped out of the global routing table shortly after the cable break.
However, traceroutes show us that after a brief break in service, traffic was again flowing via MTN Uganda's other ASN, 20294, to Infocom (AS36997) which maintained a connection to Kenya Data Networks (AS33770) which, as we mentioned earlier, survived the cable break unscathed.
Make no mistake: these cable cuts caused major headaches for Internet consumers in Uganda and Kenya. But the degree of multihoming and failover engineering on display is really rather impressive. Remember, the first major submarine cable only came ashore here in 2009! Before that, the alternatives were satellite, international dial-up or nothing.
Not all change is good, of course. When it finally came ashore, submarine fiber made satellite connectivity prohibitively expensive by comparison. All of the satellite Internet provider relationships that existed prior to 2008 are now gone, their contracts not renewed because of competition from undersea cables. Satellite backup might have done some good for the providers who were otherwise completely cut off, albeit at great expense.
The next steps in this region's evolution are already unfolding. The East African e-commerce industry is making up for lost time, and high-speed Internet access is moving inland from the cable landing stations at an amazing speed, crossing borders and tying African nations together. One can imagine that if newly independent South Sudan were to build its oil pipeline to Kenyan terminals, they, too, could go from satellite Internet to gigabit networking overnight.
Claude de Jacquelot, ICT Team Leader for the Program for International Development in Africa (PIDA), has observed that the only way to adequately hedge the risks of submarine cable breaks will be to build international terrestrial linkages that can spread the benefits of submarine landing stations to all the countries in a region, including landlocked partners like Uganda.
We predict that this will, in fact, be the case, and that the impacts of incidents like these cable cuts will become less with each passing year. African Internet is definitely headed in the right direction; by the end of 2012, PIDA estimates that Africa will be ringed by more than 90 submarine landing stations, belonging to a diverse mix of cable systems. But without more international terrestrial connectivity to provide cross-border landing station diversity, cable cuts and other landing station disasters will continue to put Internet connectivity at risk for millions of people.