Current Earth Observation (EO) data pipelines rely on centralised ground segment architectures that introduce bandwidth bottlenecks, structural latency, and single points of failure that scale unfavourably with growing constellation density. This paper presents the ESA Distributed File System (EDFS), a software prototype implementing decentralised, peer-to-peer storage and data provenance for EO satellite constellations. EDFS is built on the InterPlanetary File System (IPFS) protocol, extended with a private IPFS Cluster for replication management and a custom MQTT-based name server for content identifier resolution. The system is validated through a Kubernetes-based orbital simulation framework across five use cases and more than 100 experiment variants, sweeping constellation size (1–200 satellites), replication factor, file size, file priority, and injected hardware faults. The primary comparison baseline is TUS, a resumable point-to-point upload transport modelling the current operational downlink standard. The campaign establishes three central findings: (1) EDFS delivers faster as constellation size grows, reaching ground stations in 122–207 s at n = 100–200 compared to approximately 800 s at small constellations, by exploiting relay candidates; (2) EDFS tolerates satellite failure, recovering and delivering content from surviving replicas in scenarios where the centralised baseline delivers zero files; and (3) these benefits carry a quantifiable resource cost of approximately 15× higher CPU, RAM, and network transmit compared to the baseline. The prototype demonstrates that content-addressed, decentralised storage is architecturally feasible for EO constellation environments and provides provenance guarantees by construction.
Kubowicz et al. (Thu,) studied this question.