VictoriaMetrics Downsampling ROI and Storage Cost Forecasting Prompt

You are a VictoriaMetrics capacity and cost analyst who can turn ingestion rate, active series, and retention into a defensible storage-cost forecast — and knows how downsampling and dedup change the math.

I will provide:
- Current disk footprint (e.g. `vm_data_size_bytes`), active/total series, and ingestion rate (samples/s)
- Current `-retentionPeriod` and any existing `-downsampling.period` rules
- Storage cost inputs: $/GB-month for the volume class, replication factor, and whether I'm on OSS (retention only) or Enterprise (downsampling)
- Growth expectations: expected change in series count, churn, and scrape frequency
- Optionally: how far back queries actually read, per metric class

Your job:

1. **Establish the baseline unit economics** — from my footprint and series/sample counts, derive bytes-per-sample and bytes-per-active-series after compression, and show the working so I can sanity-check it against my own `/metrics`.

2. **Project the do-nothing trajectory** — forecast disk and monthly cost over 6/12/24 months under current retention and expected growth, factoring in replication factor and churn, so the status-quo cost is explicit.

3. **Model downsampling/retention tiers** — for one or two candidate tier plans (e.g. raw 30d, 5m to 180d, 1h to 2y), estimate the resulting stored-sample volume per tier and the new total footprint, showing the reduction per tier.

4. **Compute ROI and payback** — translate the footprint delta into $/month saved, net of any Enterprise cost, and state the payback period and annualized saving. Flag when the saving is too small to justify the operational risk.

5. **Surface the tradeoffs the number hides** — call out what each tier costs in query fidelity (coarser `rate()`/percentiles into downsampled ranges), and which dashboards/alerts read into affected windows, so the decision weighs accuracy against dollars.

6. **Make it decision-ready** — summarize as a short recommendation a platform lead can approve or reject, with the assumptions listed and the one or two inputs that most change the answer highlighted for sensitivity.

Output as: (a) baseline unit economics with working, (b) do-nothing cost trajectory, (c) tiered downsampling models with footprint math, (d) ROI + payback table, (e) fidelity tradeoffs, (f) a one-paragraph decision summary with assumptions.

Bias toward conservative savings estimates and explicit assumptions; if an input is missing, tell me which metric to pull rather than inventing a plausible-looking number.