Turbo-charging satellite Internet links into remote Pacific Islands

The research team

This is a project that involves academia, government and community organisations, NGOs and industry.

PI: Dr Ulrich Speidel (Computer Science, FoS)
u.speidel@auckland.ac.nz

AI: Dr Sathiamoorthy Manoharan (Computer Science, FoS)
s.manoharan@auckland.ac.nz

Research assistants: Wayne Reiher, Toby Tomkinson

The research

The background theme is improving Internet connectivity in the Pacific Islands. Major fibre optic cables now connect many of the larger population centres around the Pacific Islands. Many of the smaller and more remote islands, however, still use satellite Internet as their only means of digital connectivity to the outside world. For cable, the entry cost is high but the ongoing cost of operation is modest to low. For satellite, the entry cost used to be low but the ongoing cost of operation was literally astronomical. Some users / Internet Service Providers (ISPs) in the islands still pay for satellite bandwidth provisioned in the order of several hundred US$ per Megabit per second per month – many times the price commonly charged to end users in cable-connected countries. This is now in parts challenged by the availability – officially or informally – of Starlink’s low earth orbit satellite service, which presents a significant improvement for end users, if at a pretty steep price, and circumvents the ISPs almost entirely.

The old satellite links used to supply ISPs. Many of them observed that their links never hit capacity, even during busy times of the day. In many cases, utilised capacity hovered between 50% and 70% – all while end users complained of slow Internet performance. This often ended up in users finger-pointing at ISPs, and ISPs finger-pointing at satellite providers, who in turn often pointed back. The root cause behind the problem, however, is the way in which the Internet’s most common transport protocol, TCP, interacts with satellite links of the kind found in the Pacific, most of which in turn are found in this region. Our research in this area focuses on technical solutions that deal with the causes of TCP’s slowdown.

Starlink does not suffer from this problem quite as badly, but throws up new research questions in terms of reliability and scalability: Satellite cover over the Tropics is thin compared to moderate latitudes, and low earth orbit systems are a bad fit for content delivery networks, which supply over 70% of today’s downloaded data. We know that there is still scope for growth, but also that this is limited. Some of our current work is trying to determine just how much potential there is and what the real capacities of low earth orbit systems are and can be.

Fibre optic cables to islands could offer relief here with nearly unlimited capacity. But these need local ISPs to act as gateways into the local on-island networks – the very ISPs that are currently finding themselves sidelined almost entirely by Starlink. Yet these ISPs are having to invest into both the incoming cable and into an improved local infrastructure (mostly fibre) on the island(s). Will they be able to survive and master this challenge? Can their strategic importance survive the economic realities of the present day?

Other problems that have plagued Pacific connectivity over the years have been reliability, both in islands connected by cable and in those connected by satellite, and politics both at a local and geopolitical level. Since 2011, we have worked with Pacific Island ISPs and other organisations in the Pacific and beyond to document packet arrival timing in the Pacific Islands and elsewhere.In 2014/2015, we demonstrated that standard end user equipment using TCP could benefit from coded tunnels applied over the satellite links. This work was part of the PhD of Tongan student ‘Etuate Cocker, the first Tongan to obtain a PhD in Computer Science. The work was supported by PICISOC (the Pacific Island chapter of the Internet Society), ISIF Asia, Internet NZ, (then) Telecom Cook Islands, Internet Niue, and Tuvalu Telecom. ‘Etuate now works for Exclusive Networks bringing Internet solutions to a variety of Pacific Islands. We then investigated the use of coded tunnels for all traffic to and from an island formed the motivation behind the construction of the Auckland Satellite Internet Simulator Facility in Computer Science, which currently makes up most of the network lab in 303-412. This work has also been supported by ISIF Asia and Internet NZ. A number of project and Honours students (none of them Pasifika, though) have looked at the choice of off-island satellite gateway location and have developed software for the simulator. Fuli Fuli, a NZ-born Samoan MSc student, worked on performance-enhancing proxies for satellite Internet in 2018/19. In the aftermath of the eruption of Hunga Tonga Hunga Ha’apai, in cooperation with Shane Cronin and others, Ulrich Speidel provided public comment on both the cable outage and on the proposed deployment of Starlink low-earth orbit satellite service from Fiji to Tonga. He also provided private advice to decision makers in politics and industry. Education NZ Scholarship winner Wayne Reiher (from Kiribati) is currently working on measurement and performance simulation of low earth orbit networks including Starlink for his PhD. UoA Doctoral Scholar Toby Tomkinson is looking at content delivery options for low earth orbit networks.

Research questions:

• How can we improve Internet access in the Pacific Islands? Which technologies work? What is their true and theoretical capacity? How scalable are they into the future?
• How can we leverage cutting-edge developments in information theory and coding to access unutilised capacity on satellite links to islands? How can we make this technology robust and simple enough for island deployment without the need for specialist staff to travel there?

Future directions

Can we provide input into the design of low earth orbit network constellations to improve their performance and reliability, especially in tropical latitudes? It is becoming increasingly apparent that research input is needed to provide advice to the strategic development of submarine cable projects in the islands, and we are well placed to provide this. How can we build human expertise capacity in the Pacific? Islands are often just endpoints in cable networks. Can we turn them into transit points to help pay for connectivity? How do we increase the reliability of networks in the face of power cuts, natural disasters, accidents and the threat of malicious interference? Can we bring the research and education networks to the Pacific Islands that have done so much to supercharge Internet development elsewhere?