Skip to content

ADR-052: Flat DRR override for non-PowerStore storage models

Date: 2026-02-20 Status: Accepted

Context

StorePredict was originally designed exclusively for Dell PowerStore, which supports both inline deduplication and compression. Pre-sales engineers also need to size for two other Dell platforms that offer different data-reduction capabilities:

Platform Dedup Compression Effective DRR
PowerStore ✅ ✅ Per workload (from DRR.csv)
PowerFlex ❌ ✅ ~2:1 flat
PowerVault ❌ ❌ 1:1 (none)

The tool needed a way to switch between platforms without re-uploading the file.

Decision

Apply a flat DRR override at the session layer rather than adding platform-aware columns to DRR.csv or branching inside the calculation engine.

apply_storage_model() in services/drr_table.py iterates over row_data (the list of per-VM dicts stored in app.storage.tab["vm_data"]) and overwrites each row's drr key:

  • POWERVAULT → 1.0
  • POWERFLEX → 2.0
  • POWERSTORE → drr_table.get_ratio(category, subcategory) (restores from CSV)

The selected model is persisted as app.storage.tab["storage_model"] (string value of the StorageModel enum). The review page applies the stored model on every page load, so navigating away and back preserves the selection.

Consequences

  • calculation.py and report.py required zero changes — they read DRR from session data and are model-agnostic.
  • Switching model overwrites any manual per-VM DRR edits the user made. This is intentional: model selection is a global override. The drr column remains manually editable after switching for fine-tuning.
  • PowerFlex DRR=2.0 is a conservative round number based on Dell's published compression-only benchmarks. A pre-sales engineer can override individual VMs if they have more precise data.
  • Adding a fourth platform in future requires only a new enum value and a new elif branch in apply_storage_model().