Satlyt’s test validates distributed satellite computing in orbit

Satlyt, founded in 2024 by former Google and SpaceX engineer Rama Afullo, demonstrated a significant milestone earlier this month. Working with the University of Houston on a NASA-funded research project, the company completed a successful test of its software designed to distribute computational tasks among satellites in orbit. Three satellites currently carry Satlyt’s software, with the October test marking the first major validation of its core capability.

Orbital constraints that ground satellite operations

Current satellite operations face three fundamental limitations that restrict their value. Available computing resources aboard spacecraft are minimal compared to ground-based systems. Communication windows between satellites and Earth are intermittent and unpredictable. Getting useful information back to people requires either transmitting massive raw datasets or processing data onboard and returning only the results.

These constraints create inefficiency. A spacecraft observing a wildfire must transmit the full observation set rather than filtering onboard to extract only actionable intelligence for emergency responders. The delay between observation and analysis can be hours or days, reducing the value of time-sensitive information.

Satlyt’s distributed computing approach

Read also: Theranos Whistleblower Warns AI Boom Repeats Fraud Blueprint · Parents Sue OpenAI Over GPT-4o Role in Daughter’s Death · Google Releases Genie 4 World Model With Real-Time Audio Generation · Privacy-Focused AI Assistant Ollie Shuts Down November 6

Afullo designed software that enables satellites to function as a coordinated computing network rather than isolated platforms. His system automatically distributes tasks among available satellites based on computing power, communication availability, and specified priorities. Users define permissions, limits, and priorities upfront; the software makes execution decisions independently within those boundaries.

This automation solves a practical problem. Managing fleets of satellites with intermittent communication requires constant decisions about where work should run. Having the software handle these decisions eliminates manual coordination that would be impossible given communication delays and bandwidth constraints.

What the October test showed

The test operated a simulated network of four computing hosts and eight services across four users. One physical computer aboard a spacecraft executed the experiment by running this simulation and returning results to Earth. The test proved that Satlyt’s software can generate execution plans on hardware in orbit and successfully implement those plans.

The team evaluated different approaches to assigning tasks across the simulated computers while accounting for compute capacity and communication delays. One strategy outperformed others but required more processing power than alternatives, revealing a trade-off that will guide future optimization work. The software development process, onboard execution, and return of outputs all succeeded as designed.

Technical issues uncovered and next validation steps

The time-stamping problem and testing roadmap

During the test, engineers discovered that the software was recording transmission delays incorrectly. It stamped delays as arrival time, the moment data reached destination, rather than elapsed time, or how long transmission took. This distinction affects scheduling decisions in actual operations, requiring correction before deployment.

Infrastructure requirements for operational use

Satlyt has identified additional work before the software can operate on real missions. The company must test performance across varying workloads, different satellite hardware, and diverse communication scenarios. It needs secure data transfer protocols, resource scheduling algorithms that handle resource contention, and fallback procedures when satellites become unavailable. Permission from satellite operators will be required, as will usable communication links between spacecraft.

Research beyond near-Earth orbit

Broader testing lies ahead. The October experiment used simulated computers; future tests will evaluate how multiple physical computers actually exchange information in practice. Beyond low Earth orbit, Satlyt is researching applications including 4K video streaming from the Moon, funded through its NASA research agreement.

Afullo’s background and company recognition

Before founding Satlyt, Afullo worked at Google and SpaceX, where he developed expertise in both software infrastructure and aerospace systems. His experience recognized the gap between what is computationally possible on Earth and what currently happens in orbit.

The startup gained public recognition this year by reaching the top five finalists in TechCrunch’s Disrupt Battlefield competition. NASA has not yet selected Satlyt for deployment on any particular mission, though the agency is funding the current research work through the University of Houston partnership.

The infrastructure opportunity in space

Afullo described the longer-term objective: developers should be able to build applications for space without rebuilding infrastructure for each mission. This parallels how cloud computing transformed Earth-based software development. Rather than each organization building its own data center, developers deployed to managed platforms. Satlyt is pursuing the same model for orbit.

Next steps include broader validation of the software across different systems and demonstrations of computing tasks shared across separate spacecraft. NASA operates an existing network designed to move data through space even when communication links experience delays, having begun using this system earlier in the year and tested it on aircraft as well.

Source: TechCrunch