Storage Optimization
The Storage Optimization page at /finops/storage-optimization recommends and applies optimisation actions on Delta tables to reduce storage cost and speed up reads.
Optimisation actions
| Action | What it does | When to apply |
|---|---|---|
OPTIMIZE | Compact small files into larger ones | Many small files (e.g. high-frequency append) |
OPTIMIZE ZORDER BY | Cluster data by hot columns | Filter-heavy reads on specific columns |
VACUUM | Remove time-travel files past retention | Time-travel > 50% of total size |
Tier to cool | Move cold data to cheaper storage class | Last accessed > 90 days |
Tier to archive | Move very-cold data to archive class | Last accessed > 365 days |
Drop unused partition | Delete partitions with zero recent reads | Old partitions in date-partitioned table |
Reduce time-travel retention | Lower delta.deletedFileRetentionDuration | Long retention not used by consumers |
What you see
Each candidate appears with:
- Table FQN
- Recommended action
- Estimated saving (GB and $/month)
- Confidence (high / medium / low)
- Risk note — e.g. "VACUUM is irreversible; ensure no consumer relies on time-travel beyond 7 days"
Sort by estimated saving descending.
Apply
- One-click — for safe actions like
OPTIMIZEandVACUUMwithin retention - Plan — preview the change, confirm, then apply
- Schedule — defer to a low-traffic window via Scheduling
Bulk-apply across multiple tables via the Apply selected button.
Tracking
The Applied tab shows realised savings post-action, broken down by:
- Storage GB freed
- $/month saved
- Time-to-apply
- Any errors encountered
Safety
VACUUMruns withRETAIN N HOURS— never below the configured retention- Tiering changes write a manifest of moved files and can be reversed
- All actions write to Audit Trail
API
GET /finops/storage-optimization/candidates
POST /finops/storage-optimization/apply
{
"table_fqn": "...",
"action": "OPTIMIZE",
"params": { "zorder": ["customer_id"] }
}
GET /finops/storage-optimization/applied
Related
- Storage Costs — where candidates are sourced from
- Recommendations — combined optimisation feed
- Storage Metrics