1. The ship that routes power to survive

James S. A. Corey’s The Expanse returns as the recalled work because the Rocinante is constantly rerouting power, air, and water to whatever subsystem is currently keeping the crew alive. A power-constrained desktop must do the same: decide which loads matter now and which can wait.

This entry wonders about load shedding and task deferral.

2. Load shedding

Load shedding means reducing power draw when supply is insufficient. Options include:

  • Turn off non-essential payloads: cameras, processors, transmitters, and experiments that are not time-critical.
  • Reduce transmit power: lower data rate or use a lower-gain antenna.
  • Throttle compute: reduce clock speed or pause non-critical virtual machines.
  • Dim or disable heaters: accept a wider temperature range on non-critical hardware.
  • Switch to low-power modes: attitude sensors, radios, and storage all have sleep or idle states.

Shedding is not all-or-nothing. It should be tiered: first customer payloads, then non-essential platform services, then keep only attitude control, thermal survival, and communications beacon.

3. Task deferral

Task deferral means moving work to a time when more energy is available. Options include:

  • Batch data downlinks to sunlit periods when the array is generating.
  • Schedule compute-intensive jobs for the middle of sunlit time, not the end of eclipse.
  • Postpone storage maintenance and scrubbing until the battery is charging.
  • Delay propulsion maneuvers until the power budget can support them.
  • Queue customer jobs and offer predicted execution windows.

Deferral requires the platform to know both the energy forecast and the job deadlines. A job that must finish by a certain time cannot be deferred past its deadline, even if energy is cheaper later.

4. Coordinating shedding and deferral

A power constraint might trigger a sequence like:

  1. Forecast warning: notify ground and customer payloads that energy is tightening.
  2. Deferral: move scheduled non-critical jobs to predicted surplus windows.
  3. Shedding: reduce or pause remaining non-critical loads.
  4. Critical-only mode: retain only platform-safe functions.
  5. Recovery: restore loads and jobs when energy returns to margin.

5. What limits this

  • Some loads cannot be shed: attitude control, thermal survival, fault protection, and the communications beacon.
  • Customers may have contracts with minimum power guarantees.
  • Deferral requires accurate energy forecasting; bad forecasts create missed deadlines or unnecessary shedding.
  • Shedding one subsystem may increase the duty cycle of another, negating the savings.

6. What this changes

  • Load shedding and task deferral are practical autonomous responses if workloads are interruptible and priorities are defined.
  • The desktop should design payloads and services with pause, resume, and defer capabilities.
  • The next entry will ask what autonomous power budgeting needs from the desktop architecture.