Our thoughts are with the entire Tongan community during this difficult time.
The recent earthquake offshore from Tonga has brought into sharp focus the fragility of worldwide communications in the face of real-world events. For some time now, the vast majority of communication to and from Tonga has been reliant on the Internet, and there is currently a deeply concerning lack of communication.
There is a case to be made that the routine use of Voice Over IP (VoIP) and instant message apps running over the Internet that embed audio and video has removed resiliency from the system overall. There aren’t that many alternative paths to get live communications to an isolated island community like Tonga, apart from the Internet. When the single fibre link goes down, so does international telecommunications for most people.
At first, we thought it was possibly a power cut
A tweet from Jon Brewer (who actively researches Pacific telecommunications) initially proposed one possible reason — power loss at the cable station:
As Jon notes, the cable head-end station is about 1m above sea level and has ground floor infrastructure dependencies. These are very likely to have been swamped by the tsunami following the volcanic eruption.
The overall situation in Tonga isn’t entirely clear at this time, and restoring power at large to domestic infrastructure (hospitals, police, and local telecommunications) may have taken priority. Or, because of its close-to-sea nature, the cable head may have suffered significant damage.
There is now evidence of a cable cut
As Jon subsequently noted in his tweet stream, a report in the NZ Herald from yesterday stated “Fintel and TCL [Tonga Cable Ltd] undertook testing yesterday afternoon which seems to confirm a likely cable break around 37kms offshore from Tonga”.
There are now reports of more than one break in the cable.
Cable operators can determine the length of the cut using a piece of equipment called an ‘Optical Time Domain Reflectometer‘ (OTDR). An OTDR can be used to measure the distance from the meter to a break in a fibre by the time delay for the reflected light to return from the break in the cable. Even over thousands of kilometres this can be highly accurate, so if they’ve placed the loss of service along the length of the cable (which for various reasons won’t be a straight line from one point to the other: Cable laying follows the ‘best ground’, which is a function of depth and sea-bottom surface stability, as well as political boundaries and other conflicting uses of the seabed), there would be high confidence this is the cause.
Cable repair ships are stationed worldwide, and the nearest candidate is in Papua New Guinea. The NZ Herald article notes the “…likely delay to repair will be weeks, if not months”. Domestic inter-island communications will also need repair, where microwave links cannot be used.
As this event shows, natural disaster losses are a real thing: Fibres can be cut by deep-sea movements. This topic came up in a discussion at the APRICOT 2007/APNIC 23 meeting ,where an earthquake offshore from Taiwan caused significant international fibre losses. Although the fibres were notionally ‘independent’ and unrelated, they all lie within a similar corridor of suitable water offshore from southern Taiwan, and so were subject to the same deep-sea earthquake. The notional redundancy of ‘different fibre’ disappears when they all lie in the same place.
However, we still don’t know all the effects of the Tongan volcano. What we have seen so graphically on the satellite photos was the huge explosive power — primarily about ejected material — not ground movements that can cause fibre breakage. Unfortunately, even in this case, there will have been sufficient undersea movement that cable loss is likely.
What about backup links via satellite?
There is pre-fibre satellite telecommunications infrastructure in Tonga. Ground stations for geosynchronous satellites exist, but like other infrastructure is highly dependent on local power. Tonga was connected to fibre communications with Fiji in 2013. Before then, it had been entirely reliant on satellite communications for Internet and data services in general, which were subject to cyclone damage and loss of service from time to time, as well as speed and delay issues. Geosynchronous communication satellites impose higher delays and are much harder to upgrade to higher speeds than planned at launch, where fibre communications can often be sped up by changing the wavelength of the transmissions down the same glass inside the fibre cable.
Satellite communications will also be affected by dust clouds. Ash forms a layer on the dish and causes attenuation of the signal. The primary consequence of this kind of interference is a drop in bandwidth. So that higher (stronger) forms of error-correcting codes can be used, less bandwidth is available to be shared by consumers of the link, because of the increase in overheads to ensure reliable communications. Aside from this, more people are looking to use the link than it has capacity for, because the fibre increased expectations of bandwidth availability. This is a strong instance of Murphy’s Law — when you maximize use of the satellite fallback path, and have MORE people using it, FEWER people can get what they want. It’s very likely congestion will affect communications to the Tongan community at large for some time.
Lower orbit satellite telecommunications such as Starlink will be similarly affected, but may not currently be offering service over this region due to economic, regulatory, and infrastructure considerations. However, it may offer hope for rapid recovery and deployment.
Similar to Starlink but with much lower bandwidth, a small number of satellite portable phones such as the Iridium system might also be available.
Defence in depth is good but sometimes ‘force majeure’ applies
The general principle here may be an instance of a process we use in the online backup community called 3-2-1 strategy: Three forms of backup, two of them ‘live’ but different, and one offline. If we apply this to the situation at hand, Tonga actually had two live methods in deployment: The fibre and the older geosynchronous satellite communications both existed and both were live (or near-live, assuming the satellite communications haven’t been decommissioned but are not in active use). It’s arguable that a second fibre path would be beneficial but it needs to be noted that fibre is not cheap. The Tonga fibre is 827 kms long, running to Fiji, and there might not be good affordable alternative places to connect to that would provide resiliency. There is precedent, with a fibre cut causing outage in 2019.
In reality, you can’t always cover for all eventualities. Some events may be too large to prevent loss of all your backup choices.
This is an emerging story and the tech lists and online resources will continue to explore exactly what’s going on in Tongan telecommunications for some time. On behalf of everyone at APNIC, we all hope for a speedy recovery for those affected by this disaster.
The views expressed by the authors of this blog are their own and do not necessarily reflect the views of APNIC. Please note a Code of Conduct applies to this blog.