Skip to content

Handover

B. Instant Handover Re-Optimization

When a telescope goes FAULTED during a critical SSA track, the scheduler executes the Replication/Primary-Backup strategy:

  1. Stop Track: The central system sees the FAULTED heartbeat and immediately marks the current track as failed for that node.

  2. Backup Trigger: The scheduler immediately checks for any telescope currently in the HANDOVER WINDOW (the overlap time where a second telescope could also see the target).

  3. New Assignment: If no overlap is possible, the scheduler immediately searches the global database for the next telescope that will acquire the target, often scheduling a small buffer of backup images to ensure the gap is covered.

Priority Queue Design and Reliability Weighting (from massive info dump.md)

Priority Queue Structure

The scheduler uses a simple priority queue — avoid Integer Linear Programming (ILP) which is too brittle for a volunteer-grade system.

  1. Alerts / ToO (Highest Priority): Gravitational wave follow-up, rapid transients. Time-critical; pre-empts all other observations.
  2. SSA Tracks (High Priority): Time-sensitive satellite passes and NEO tracks; require handover logic.
  3. Deep Sky (Lowest Priority): Long, non-time-critical integration requests; can be deferred.

Reliability Weighting

Introduce a Reliability Score (0–1) for each node. The scheduler prefers higher-reliability nodes when multiple sites can observe the same target. The score is dynamic — updated based on actual success/failure rates over a rolling window.

  • High reliability: consistent dry-climate site, experienced operator, recent track record of completing tasks
  • Low reliability: sporadic humid-climate site, beginner operator, frequent cloud-out failures

This prevents low-quality but technically "available" nodes from being assigned to time-critical tasks.

Data Heterogeneity Management

For deep-sky stacking jobs: - Group telescopes into "families" based on similar FoV and pixel scale. Only stack images from similar-family telescopes in a single session to maintain consistent final image scale. - Automatically assign low weight to images with poor FWHM or low SNR during co-addition, regardless of contributor identity. Bad data should never degrade the final product.

See also: Telescope Capability Vectors.md for the full vector-based matching approach.