Software designed to distribute computing tasks among satellites has generated an operational schedule on a real onboard computer in space. It is being developed by Satlyt, a startup that wants to turn existing satellites into a data-processing network. A network connecting multiple spacecraft does not yet exist. The test also revealed that the software incorrectly records communication delays.
A Simulated Network in Orbit
The test took place in October after 13 months of collaboration with the University of Houston as part of a NASA-funded study. A network of four computing nodes, four simulated users, and eight services was modeled on a single physical computer aboard a spacecraft. The software determined where to run each task based on the nodes’ available computing capacity and connectivity. The results were returned to Earth.
Satlyt founder and CEO Rama Afullo discussed the test with TechCrunch. One scheduling approach performed best, although it required more computing power than the others. The test established that the software can generate schedules on orbital hardware, but data exchange between multiple separate computers still needs to be tested.
Why Satellites Need Shared Computing
Satellites collect increasing amounts of data, but they can transmit it to Earth only when passing over a ground station. In low Earth orbit, a communication session with a single station typically lasts five to ten minutes per orbit. Processing the data onboard would mean that only the final result needs to be transmitted.
Rama Afullo illustrates this with the example of a forest fire. Once a satellite detects it, software like this could help deliver the necessary information much faster.
Onboard computers have limited resources, so the startup wants to connect several spacecraft into a network where tasks are assigned to whichever node has available capacity and connectivity. Operators would specify only permissions, priorities, and operating limits, while the software would make decisions within those boundaries. We previously reported on the company’s plans to bring satellites together into a computing cloud.
An Error in Recording Delays
The test also exposed a flaw. The software recorded data transmission latency as the time a message arrived, rather than the elapsed time between sending and receiving it. For communications in Earth orbit, the distinction may seem minor. However, a signal takes about 1.3 seconds to reach the Moon and between 3 and 22 minutes to reach Mars, depending on the planets’ relative positions. Software that confuses these values would assign tasks based on an inaccurate understanding of connection quality.
Nevertheless, the experiment was not a failure. According to Rama Afullo, development, onboard execution, and the return of results were successful. The company will next test the software with different workloads, hardware, and communication conditions.
From the Moon to Multiple Spacecraft
Satlyt’s software is already running on three satellites, although NASA has not yet selected it for any mission. The collaboration with the agency extends beyond Earth orbit.
The startup received funding through the STTR program, which supports joint projects between small businesses and universities. According to the project description in NASA’s archive, the aim is to ensure that high-definition video from the Moon reaches an operator even when connections between individual parts of the network are interrupted. This will involve autonomous caching. Copies of the video will be stored at intermediate network nodes, preventing data loss during connection outages and allowing transmission to resume as soon as the next link becomes available. Rama Afullo specifies that the video will be in 4K.
The company identifies its next step as a demonstration of shared computing between separate spacecraft. This will require functioning communication links, agreement from satellite operators, secure data transmission, and the ability to recover when a node drops out of the network.
