Since the round-trip latencies are now below 200ms to locations in the US, this cannot be going over satellite.
Mystery Cable Activated in Cuba
Recent developments captured in a new blog here. (January 22, 2013)
|In February 2011, the first submarine cable connecting the island nation of Cuba to the global internet (by way of Venezuela) landed on Siboney beach, Santiago de Cuba. In the two years since, the fate of the cable has been a mystery for Cuba observers. In the past week, our global monitoring system has picked up indications that this cable has finally been activated, although in a rather curious way, as we explain below.|
Connecting Cuba to the Internet
In 2007, state-owned telecommunications companies from Cuba and Venezuela joined forces to build a submarine cable between the two Caribbean nations, linking Cuba directly to the global Internet and allowing it to end its reliance on satellite-based Internet services. At least that was the hope. The cable was named the “Alternativa Bolivariana para los Pueblos de nuestra América” or ALBA-1 for short.
Originally planned to be completed in 2009, the project hit delay after delay, until construction was finally completed in early 2011. However, despite the announcement of its completion, Cuba’s Internet has still limped along on high-latency satellite service via three different Internet service providers. That is, until last Monday when we noticed that Spanish telecom giant Telefonica began service to Empresa de Telecomunicaciones de Cuba S.A. (ETECSA), the state telecom of Cuba.
To underscore the significance of this development, we look back at how ETECSA obtained Internet service over the past 6 years. During this period, we see the same three satellite providers, although the routes carried over each has varied considerably over the years (below left). However, zooming in on the last couple of months, we see the entrance of Telefonica in our routing data as of last week (below right).
We believe it is likely that Telefonica’s service to ETECSA is, either by design or misconfiguration, using its new cable asymmetrically (i.e., for traffic in only one direction), similar to the situation we observed in Lebanon in 2011. In such a configuration, ETECSA enjoys greater bandwidth and lower latencies (along the submarine cable) when receiving Internet traffic but continues to use satellite services for sending traffic. While the activation of the ALBA-1 cable may be a good first step to providing ETECSA a better link to the Internet, the lack of widespread public access to Internet service throughout the island will likely continue.
On the same day last week that we saw the first evidence of the ALBA-1 cable, Cuba eliminated the requirement of an exit visa for its people to travel outside the country. Could these two developments be part of a greater trend towards a freer and more open Cuba?
Nice article! A pity that you could not stay out of politics and had to bring the unrelated (for a Rennesys blog) topic of travel. "Could these two developments be part of a greater trend towards a freer and more open Cuba?" You could save that part of the analysis for other types of blogs I guess.
From Simon Lloyd, head of communications, Telefonica UK: "I can confirm that Telefónica has no involvement with the submarine cable ALBA-1, which links Cuba with Venezuela. As a global telco operator, Telefónica provides international wholesale services to local operators and companies around the world. Telefónica is not alone among global telco operators in providing services to Empresa de Telecomunicaciones de Cuba S.A. (English: Telecommunications Company of Cuba; ETECSA)." Renesys did not assert Telefonica had anything to do with the ALBA-1 cable itself. Just that ETECSA is using the cable to reach Telefonica service in Venezuela, where they have a significant presence. The last sentence confirms that Telefonica (like 3 other providers) is providing service to ETECSA.
Hi Pedro, Thanks for your question! I do not see Internet traffic traversing Cable and Wireless's infrastructure to reach Cuba, just Telefonica and the three satellite providers. In addition, we have observed no recent change traffic to Jamaica.
At this moment Cuba prioritizes the use of internet in goverment,businees,universities,schools,libraries and other social interest, not for personal use. Hope that in near future the access expanding to all areas.
Hi Doug, At the end of cable installation was mentioned that Cable and Wireless (AS1273) from UK was selected as the operator of this cable. Based on this, we can guess that the main gateway for Cuba would be Jamaica instead of Venezuela. Has been any change in C&W traffic?
it could be simpler: they are dropping some dual satellite hops and therefore you are seeing a latency drop.. and that's it.
it seems to be an improvement on Cuban Internet provider Services, switching from an unstable and low speed satellite internet connection, into submarine wire connection. but the ironic of this is, in cuba the internet is for tourist use only. is not a benefit for all the cubans living in the island.
George, TCP delivers on its reliability features by ACK'ing every packet such that if any single packet goes missing it can be retransmitted without the application layer worrying about corrupt data. You can tweak some aspects of the protocol to increase the number of "in-flight" packets waiting to be ACK'd but ultimately it slows the whole process down.
Thanks Doug. Interesting post. Am I correct in thinking then, that for "browsing" this connection would be mediocre, but for constant transfers once the initial ACKs have been sent, it would be much faster?
Definitely something smell fishy here... Born in Cuba and knowing the government tentacles is very probable they are testing filters (firewalls) first, and then will expanding connection to limited areas... The worst scenario for Castro is every people in the Cuban net accessing videos and info from the outside world
Video streaming via UDP is not affected by this. As a former technician there myself I remember how we would reduce download speeds or even disconnect some users so that our torrents could get a bit more bandwidth overnight. If this is not what's happening I would go with the firewall theory. The diurnal pattern does not apply too much in Cuba since many people share time slots online and the best time to be connected is late at night. I used to live on this "vampire time" for some time, basically you sleep during the day and connect at night.
If that is their logic, it doesn't make much sense -- RTT impacts TCP throughput. With a high RTT, they'll still experience crummy download speeds. See this (just the first function is enough, the rest is just getting tighter precision...): http://www.slac.stanford.edu/comp/net/wan-mon/thru-vs-loss.html
Cuban technicians are more interested in high speed downloads. Most of them, if not all, run underground businesses of supplying movies, documentaries and pirate software to the black market. They have families to feed on $20 a month like most Cubans. Uploads are really not profitable for them and it can get them in trouble if something important leaks from the island so they would rather keep them under a cap at least for now. I bet a lot of this traffic is from torrents or ssl newsgroups if they are smart.
Maybe they are just making some tests... However, if you decide to support the firewall theory, would be important to consider that patterns of Internet use in Cuba are very peculiar... Slowness caused by diurnal use is not a reliable indicator...
This is a very good theory. However, I suspect this is not the case. In countries where we see latencies are impacted by censorship regimes, we often see a diurnal pattern in latencies. This is due to traffic slowing during busy times when everyone is awake and using the Internet, and the censorship software is struggling to keep up. When looking at the distributions of these latencies over time, I see no diurnal pattern and instead see what appears to be a hard floor that latencies cannot go below. This suggests some physical factor that can never be overcome. As stated in the blog, the latency distribution appears very similar to cases of asymmetric satellite we have observed in the past. Thanks for the question!