Files
Files lets you upload data directly to Saiku Cloud rather than connecting a warehouse. Drop a Parquet, CSV, or JSON file onto the page; once it’s uploaded the Schema designer treats it as a source just like a database table — profile it, propose a cube, save, query.
It’s the right surface for prototyping with a sample export, for one-off analyses against data that doesn’t live in a warehouse yet, or for reference tables small enough to ride alongside your main data. For production workloads on hundreds of GB, you’ll want Connections instead — file storage is capped by your plan and CSV doesn’t scan as fast as a columnar warehouse.
Uploading
Drag a file onto the page or click Upload to pick one. The upload runs in the background; once it lands, the file shows in the list with its size, detected format, and upload date.
Supported formats:
- Parquet (
.parquet) — columnar, by far the most efficient. Saiku reads only the columns your query touches, so a multi-GB Parquet file scales well even with the shared engine. - CSV (
.csv) — with a header row, comma- or tab-delimited. - JSON (
.json) — either newline-delimited records (JSONL) or a single array.
Building a cube from a file
- Go to the Schema designer.
- Pick File as the source type and select an uploaded file from the dropdown.
- The designer profiles the file the same way it would profile a warehouse table.
- Continue with the normal designer flow.
Quotas
| Tier | Storage |
|---|---|
| Starter | 5 GB |
| Team | 50 GB |
| Business | 500 GB |
| Enterprise | 5 TB+ (custom) |
The page header shows your current usage. Once you hit the quota, new uploads fail until you delete files or upgrade. Existing queries against already-uploaded files continue to work.
Deleting
Trash icon on the row, confirm. The file is purged from storage
immediately and any cubes pointing at it return a source not found error until you re-upload or repoint them.
When to prefer a warehouse instead
- Data changes daily or hourly. A file upload is a snapshot. If the underlying data is moving, point at a warehouse so cubes always read fresh.
- Your dataset is over 5 GB and you’re on Starter. You’ll hit the quota fast. Either upgrade or move the data into a warehouse.
- Multiple cubes need to join against the same dataset. A warehouse table can be joined by many cubes; a file upload is read once per cube.
Related
- Schema designer — uses uploaded files as a source.
- Connections — the warehouse alternative.
- Usage — watch your storage trend.