Operational Co-Flow Optimization of a Serviceable Space Data Center: Binding Capacities and Marginal Values at Fixed Designs

This paper formulates one operational day of a space data center, which processes client data onboard for a ground network, as a cross-domain co-flow program on one time-expanded network with cyclic boundary conditions, solves it as a linear program at ten-minute resolution, and reads from its duals the marginal value of each sized subsystem per kilogram launched, audited against finite differences. A daily aggregate is derived and its range of validity measured, and a threshold rule derived from the optimum is replayed with exact radiator physics until periodic. The study covers four reference designs, eight campaign years and three levels of client demand. The design fixes which capacities can bind; demand and technology generation decide which one binds and which workload is marginal, and the price of the binding capacity is the value of that marginal workload per unit of capacity: a kilogram of radiator is worth 71 GBeq per day while the imaging workload is marginal and 373 once the radar workload is. The rule retains at least 95% of the program’s value in ten of twelve cases and loses 13% where compute, heat rejection and downlink saturate together, and 7% where the eclipse energy budget binds.