Juan A. Fraire

Researcher in Space Networking and Informatics

Inria CONICET Saarland University

Press kit

For journalists writing about orbital data centres, satellite constellations or the internet beyond Earth: a bio in two lengths, a photo, results with their sources, and how to reach me.

  • Portrait of Juan A. Fraire, in colour
Download in colour (510 KB)
Download in black and white (650 KB)
JPG, 1506 × 2148 px

Contact

  • Emailjuan.fraire@inria.fr
  • Based inLyon, France (Central European Time)
  • Interviews inSpanish (native), English and French
  • AffiliationsInria · CONICET–UNC · Saarland University

Write to me

Bio

Short, about 50 words

Juan A. Fraire is a researcher at Inria in Lyon, France, and a guest researcher and professor at CONICET–UNC in Argentina and at Saarland University in Germany. He works on networks and computing in space. He chairs the IRTF SPACE Research Group, co-founded and chairs the STINT workshop, and has published more than 160 papers.

Long, about 150 words

Juan A. Fraire is a researcher at Inria, the French national research institute for digital science and technology, in the Agora team in Lyon. He is also a guest researcher and professor at CONICET and the National University of Córdoba (UNC) in Argentina, and at Saarland University in Germany. His work joins telecommunications and computer science to build networks beyond Earth: delay-tolerant networks for deep space, satellite links for Internet-of-Things devices, routing across constellations of thousands of satellites, and computing in orbit, including its energy and carbon costs. His contact-graph routing code runs in ION and HDTN, NASA's networking software for space. He chairs the SPACE Research Group of the Internet Research Task Force (IRTF), co-founded the Space-Terrestrial Internetworking Workshop (STINT) in 2014 and still chairs it, and sits on the board of IPNSIG. He has published more than 160 papers and supervised 16 doctoral students, seven of whom have completed their PhD.

What I can talk about

Results on orbital computing, with sources

From papers with colleagues at Saarland University, Nanjing University and the University of Genoa. Each figure links to the paper and the section it comes from. More on the orbital computing page.

Where it can work

  • Distance to the ground. A cloud in orbit answers faster than one on the ground when the user is more than about 5,000 km from the nearest gateway. Under 2,000 km, the ground wins. ASMS/SPSC 2025, §V-A
  • Shrinking data before sending it. Processing in orbit saves carbon when it cuts the data enough: with one inter-satellite hop, down to 50% of the raw size on Falcon 9, or 70% on Starship. Dirty Bits, §4.3
  • Against an average data centre. Within its first one or two years in orbit, a node emits less carbon per GPU-hour than the global-average data centre on Earth (about 477 g of CO₂e). Dark Clouds, LEO-NET 2026, Fig. 1

What it costs

  • Heat. Every watt a computer draws ends up as heat, and in vacuum it can only be radiated away. For a 1 kW node, the radiator weighs 30.5 kg of the 44.3 kg modelled; the computer itself, 3.2 kg. Dark Clouds, §4 Scaled to the 250 kW that SpaceX quotes for each of its Star Mind spacecraft, the same ISS-based density gives about 760 m² and 7.6 tonnes of radiator, the heat load of some 18 ISS radiator panels. Dark Clouds, §3
  • Launch and re-entry. Each kilogram costs about 52 kg of CO₂e on Falcon 9 and 34 on Starship. Re-entry adds about half again on top of the launch. Dirty Bits, §3.1
  • No repair. Spares have to fly from day one. One cold spare of the computer adds 40% to the carbon per GPU-hour on Starship, and 34% on Falcon 9, for a three-year mission. Dark Clouds, abstract
  • Against a clean data centre. Matching a data centre on a clean grid (about 107 g per GPU-hour) takes at least six years in orbit with one spare on Starship, or nine years with two spares on Falcon 9. Dark Clouds, §4

Carbon figures count manufacturing, launch and re-entry of the modelled components; once in orbit the node runs on sunlight. The satellite bus (structure, attitude control, avionics) is left out, as in the papers; including it would add mass.

Results on space networks, with sources

From papers written with students and colleagues. Each figure links to the paper and the part it comes from.

01Delay-tolerant networks

  • Tested in orbit. In December 2020 and May 2021, ESA's OPS-SAT satellite stored and forwarded a web request and its reply between a simulated offline site and a site with internet access. The paper reports it as the first successful test in space of version 7 of the Bundle Protocol, the standard for this kind of store-and-forward networking. Ring Road Networks, IEEE Comm. Magazine 2022, In-Orbit Validation In simulations of ten remote ground stations, the average delivery time fell from 7.7 hours with one small satellite to 1.3 hours with 20. Ring Road Networks, Fig. 2b
  • Laser links at Mars. In 24-hour simulations of Mars orbiters with one laser each, relaying science data to Earth, a scheduler that plans every link around the flow of data toward Earth could send back 1.5 to 2 times as much data as an earlier automatic method. Mars laser scheduling, IEEE WiSEE 2024, Fig. 4a

02Direct-to-satellite IoT

  • Nine satellites for the whole planet. If devices can go up to two hours without hearing from a satellite, the longest the LoRaWAN standard expects them to keep their timing, nine satellites in well-chosen orbits can reach them anywhere between 80° north and 80° south. Keeping every point in view at all times would take 88. This counts coverage only, for orbits at 700 km. Sparse Constellations, IEEE TAES 2022, §V-A, Fig. 6
  • Waiting longer means fewer satellites. To serve every device across Africa, or across Europe, the designs in the study need 12 satellites if devices must hear from one at least once an hour, 4 if they can wait two hours, and 3 if they can wait about three. Sparse Constellations, IEEE TAES 2022, §V-B, Fig. 12

03Mega-constellations

  • Satellites to reach every under-served address. About 4.7 million US addresses have no provider offering broadband at 100 Mbps down and 20 Mbps up. Serving all of them without overselling capacity beyond the 20-to-1 ratio that US regulators allow fixed wireless providers would take Starlink over 40,000 satellites, more than five times the roughly 8,000 it had at the time of the study. The authors present this as a lower bound. Anyone, Anywhere, HotNets 2025, §3, Table 2
  • Who can pay for it. At $120 a month, Starlink's residential plan would cost more than 2% of household income, a common affordability threshold, at about 3.5 million of those 4.7 million addresses, and at nearly 3 million even with the federal Lifeline discount. Comparable cable plans at $40 to $50 a month are affordable at more than 99.99% of them. The study uses each county's median income. Anyone, Anywhere, HotNets 2025, §4, Fig. 4
  • Getting lasers to lock on. Before a laser link between two satellites can carry data, both must turn to face each other and find the beam. In simulations of two 12 kg satellites, that left only 72% of a contact for data on the harder of two passes. Turning early and exchanging satellite-navigation positions over a radio link raised the share to over 99%, and cut the memory needed to hold waiting data at 100 Mbps from 4 GB to 75 MB. RF-Assisted Compensation, IEEE OJ-COMS 2026, §IV-B, Tables 8 and 9

Apart from the OPS-SAT test, these are simulation results, under the assumptions set out in each paper.

Work in use

More on the software page.

Media and outreach