Storage Tiering Service  ·  View 04 of 31  ·  2 · People and journeys

Journey: Demote Idle Documents to Cold

The economics lead takes a policy from a merge request to every eligible object, and learns at the first ring that the corpus disagrees with the model.

Editable source SVG draw.io All views
Storage economics lead policy v42 Goal — Move documents idle for 90 days to cold, net positive Trigger — Quarterly bill 11% over plan; warm tier at 81% Done when — Full rollout, realised delta within 10% of shadow 1 · Draft 2 · Shadow 3 · Dry run 4 · First ring ◆ moment of truth 5 · Widen 6 · Score What they do Writes policy in CUE Runs shadow on corpus Dry run on 1% sample Approves 0.1% ring Reads realised delta Scores at day 90 System behind it Policy repo Shadow runner Break-even evaluator Ring controller Cost ledger Decision records How it feels Confident Unsure Anxious Where it hurts Which window per class? Projects 38%, not 45% Re-read rate 4.1% Early deletes creeping What the platform gives Ladder does the arithmetic Bill delta, 0 bytes moved Losing objects named Ring halts itself Revert free for 24 h Predicted vs realised Journey — Economics Lead Demotes Idle Documents to Cold v 1.0 · owner Storage Platform Architecture · date 2026-09

The worst moment

  • The first ring. At 0.1% of eligible objects, 4.1% of what was demoted is read again by a promoting class within 14 days, against a 1.5% gate. Almost all of it is quarter-end finance workbooks: idle for 97 days and due to be reopened in the next fortnight. The policy used the 90-day document window; that workspace class should have been declared periodic.
  • Without an automatic gate the lead would be deciding, alone and at speed, whether a number is noise. With one, the ring halts itself and the question becomes which object class was mis-declared.

What the architecture owes this journey

  • Shadow mode over the whole corpus, so the first number the lead sees is a projected bill delta, not a moved byte (ADR-22).
  • A dry run that really copies a 1% stratified sample to scratch space, so request cost and throughput are measured rather than estimated.
  • A 24-hour release hold, so reversing a bad ring is a catalogue pointer flip rather than a retrieval (ADR-03).

Numbers

  • Rings at 0.1%, 1%, 10% and 100% of eligible objects, each held for at least 72 hours. The gate reads re-read rate for immediate destinations, wrong-tier rate for delayed ones, and the charge ratio for both. No policy may move more than 2% of objects or 1.5 PB in any seven-day window.