Skip to content

ADR-076: Remove compute sizing, redirect to PreSizion

Date: 2026-03-15 Status: Accepted Supersedes: ADR-062 (now Superseded)

Context

StorePredict's /compute page provided ESXi host count recommendations based on uploaded VM data: vCPU/RAM aggregation, N+1 HA sizing, vMSC stretch cluster splits, and active/passive DR modeling. This functionality has been fully superseded by PreSizion, a dedicated sizing tool with broader compute, storage, and network sizing capabilities.

Maintaining parallel compute sizing logic in two projects creates drift risk and duplicates engineering effort.

Decision

Remove all compute sizing logic from StorePredict:

  • Deleted: pipeline/compute_sizing.py, data/compute_presets.csv, tests/test_compute_sizing.py
  • Cleaned: compute config keys from session_archive.py, tooltip i18n keys, COMPUTE_PRESETS_CSV_PATH from config.py
  • Replaced: /compute page now displays a card with links to PreSizion's web app and GitHub repository

The /compute route and nav link are preserved so users discover the redirect rather than hitting a 404.

Consequences

  • Positive: Single source of truth for compute sizing (PreSizion).
  • Positive: ~500 lines of pipeline code and ~160 lines of UI code removed.
  • Positive: Simpler session archive format (no compute keys to serialize).
  • Negative: Users must switch to an external tool for compute sizing.
  • Negative: Session archives created with schema v1 that contain compute keys will silently ignore those keys on restore (harmless — the data is unused).