Before and after Cuba's shift to the ALBA 1 undersea cable

Northwestern researchers Zachary S. Bischof, John P. Rula and Fabian E. Bustamante have published In and out of Cuba, a paper characterizing Cuba's connectivity.

They gathered data during March and April 2015 and found that traffic going out of Cuba typically traveled through the ALBA-1 cable, but traffic coming into Cuba was often routed over satellite links, adding about 200 ms to round trip times (RTTs).

The data was gathered using RIPE Atlas probes, small, USB-powered hardware devices that hosts attach to an Ethernet port on their router. There are 8,771 of these probes around the world, including one in Havana. Upstream results from that probe may or may not have been representative of the nation as a whole.

While the study results were accurate last Spring, they do not reflect the current situation. Last July, Doug Madory, Director of Internet Analysis at Dyn Research alerted me to the fact that Cuba's international traffic had largely shifted to the undersea cable.

I forwarded the Northwestern paper to Doug and he sent me the following graphs, showing traffic to both of Cuba's autonomous network operators for March 30, 2015 and October 30, 2015:

In March, there was considerable satellite traffic from NewCom and Intelsat.

By October, there was little satellite traffic.

As mentioned above, the Atlas probe may not have been representative. The graph on the left side of the following figure shows traffic to the IP address range the Atlas probe resides in. During the time the Northwestern team gathered their data, inbound traffic was almost entirely via satellite (Intelsat or NewCom), but it shifted subsequently. The graph on the right shows that the majority of traffic coming to representative destinations of the Cuban ISP ETECSA was routed over the cable during the study period. In July, it nearly all shifted to the cable, where it is today.

Most traffic to the probe shifted to cable in June (left). ETECSA received relatively
little satellite traffic during the study period and essentially none after June (right).

The analysis by the Northwestern team is thorough and insightful, but it seems they have been caught by academic publishing delays. They say this was just the start of ongoing study of the Cuban Internet and I am looking forward to seeing their future work.

Finally, today's Internet access and speeds in Cuba are very poor but, given today's low Internet penetration, domestic infrastructure, not international capacity is the key constraint.