Usage
Overview
The Usage page lets you check organization-wide VM, storage, and network resource usage along with accumulated cost. Usage is aggregated by the second, and you can filter by period, resource type, and pricing type to analyze cost patterns.
Billing.Billing.READpermission. All four built-in roles have it.
How to open it
Go to Billing > Usage.
Page layout
A summary card at the top, a time-series table at the bottom.
| Area | Contents |
|---|---|
| Total Usage | Total cost for the selected period |
| Period | Start date – end date |
| Usage table | Cost per resource type, daily or weekly |
Table columns:
| Column | Meaning |
|---|---|
| Pricing type | On-demand / Spot (reserved is separate — see below). Spot applies only to Elice-integrated organizations |
| VM | VM runtime × rate |
| Block storage | Size × time × rate |
| Block storage (DR) | DR block storage size × time × rate |
| Object storage | Stored size × retention time |
| Parallel file system | Size × time |
| Public IP | Allocation time |
| Load balancer | Time in use after activation × rate |
| Total | Row total |
Usage and billing details for reserved capacity are not available here. Check with sales or the support channel. For how to find reserved usage, see the Reserved resources guide.
Filters
Narrow the view with the filters above the table.
- Time Granularity: Daily / Weekly
- Resource Type: All / Virtual Machine / Block Storage / Block Storage (DR) / Object Storage / Public IP / Parallel File System / Load Balancer
- Pricing Type: All / On-Demand / Spot
- Period: pick the month to view
To reset all filters at once, click Clear filters.
Cost optimization checklist
- Stop VMs you aren't using — note that block storage and public IP still bill
- Delete block storage you don't need any more (back up via snapshot first if needed)
- Release unused public IPs
- Delete unused load balancers: they keep billing while activated even with no traffic
- (Elice-integrated organizations) Move interruption-tolerant ML training to spot VMs: discounted rates (discount varies)
- Auto-prune old snapshots / automatic backups with snapshot scheduler retention
- Move long-term archival data from block storage to object storage
- For workloads running 3+ months continuously, switch to reserved — contact sales
Stopping halts only compute charges. Block storage and public IPs are billed until you delete them. For data you won't touch for a while, move it to object storage and then delete the block storage.
Cost alerts
If you can't review the Usage page daily, set up alerts for cost/usage thresholds.
| Scenario | Example alert |
|---|---|
| Idle GPU | GPU utilization < 5% for 1 hour → email |
| Disk-full risk | Block storage utilization > 85% → email |
| Training completed | GPU utilization < 5% for 10 minutes → email |
Next steps
- Pricing model: on-demand vs reserved vs spot
- Official price list: hourly rates by instance type
- Alerts: notify on usage spikes
- Audit log: trace cost spikes (who created the large instance)