The NASA Jet Propulsion Laboratory (JPL) is developing the Mission Data System (MDS) for potential use in future space missions, where it is expected to reduce the complexity and effort required to produce mission and ground-support software, while also enabling greater autonomy for remotely deployed systems, including future planetary rovers. The MDS software architecture includes two levels of processor scheduling: a high-level planner and a low-level execution manager. JPL and MITRE are investigating the use of Time-Utility Functions to manage low-level scheduling, since they promise to enable adaptive, short-term processor scheduling based on mission utility. Recent development work on MDS has focused on organizing the software so that low-level processor scheduling will be most effective. This paper describes the major organizing principles that have shaped this work.