> ## Documentation Index
> Fetch the complete documentation index at: https://docs.rafftechnologies.com/llms.txt
> Use this file to discover all available pages before exploring further.
# Object Storage
> S3-compatible object storage on the Raff Platform — NVMe SSD buckets, AWS-SDK-compatible, with the same regional endpoint and SigV4 auth model as AWS S3
Updated May 8, 2026
Object Storage is **S3-compatible** storage for unstructured data — media files, application backups, logs, static assets, machine-learning artifacts, anything you'd normally throw at an S3 bucket. Use any S3 SDK or tool against the Raff endpoint, organize objects into buckets, and serve them publicly or privately.
Today there is **one storage class — NVMe SSD** — tuned for high-performance access. A cost-effective HDD class is **coming**.
## S3 endpoint
Every bucket is reachable via the Raff S3-compatible endpoint:
```
https://.s3.raffusercloud.com
```
Point any AWS-S3 SDK at that endpoint with your access keys and you're connected — no Raff-specific client library needed.
## Object Storage vs Volumes — picking the right one
Both store data; they're built for different access patterns.
| | **Object Storage** | **[Volumes](/products/store/volumes)** |
| ------------------------- | ---------------------------------------------------------------------- | ------------------------------------------------------------------ |
| **Access protocol** | HTTP / HTTPS via S3 API | Block device via virtio-blk (looks like a local SSD inside the VM) |
| **Mounting** | Not mounted — accessed by SDK / `s3` URL | Mounted into the filesystem like a regular disk |
| **Concurrency** | Many clients read/write the same bucket simultaneously | One VM at a time (single-attach) |
| **Latency profile** | Tens of ms typical (network round-trip + S3 op) | Sub-millisecond (local NVMe) |
| **Throughput per object** | High; multipart upload scales horizontally | Highest — direct disk |
| **Pricing** | \$0.07/GB/mo (\$7/mo first 100 GB included), egress free up to 1 TB/mo | \$0.05/GiB/mo flat |
| **Best for** | Unstructured data, lots of objects, many readers | Files an OS / database needs to read locally at low latency |
### When to use Object Storage
* **User-uploaded files** — avatars, photos, documents, generated PDFs
* **Application backups** — scheduled DB dumps, log archives, app artifacts
* **Static asset hosting** — site images, JS/CSS bundles, video thumbnails (point a CDN at the bucket)
* **Machine-learning artifacts** — datasets, model weights, training checkpoints
* **Customer file storage** — anything customers will retrieve via signed URLs
* **Anything you'd put in AWS S3** — it works the same way
### When to use a Volume instead
* **Database storage** — Postgres / MySQL / MongoDB data directory needs sub-ms disk access
* **Application file storage** — apps that mmap files, read configs from disk, or expect a local filesystem
* **VM extra capacity** — you need more space than the plan's base disk and want it to look like a regular disk to the OS
* **Anything that requires `O_DIRECT` / `fsync` performance** — block-level I/O Object Storage can't match
### Common mistake
**Don't run a database on Object Storage.** S3-API access has too much per-call overhead for the millions of small reads/writes a database makes. Use a Volume for the data directory and (optionally) Object Storage for backups of that database.
***
## Most viewed
Create your first bucket from the dashboard.
Drag-and-drop upload via the dashboard.
Get S3 credentials for SDK access.
Connect from your application.
Storage, requests, and egress per bucket.
***
## Browse
Buckets, uploads, access keys, SDK use.
What S3-compatibility means in practice.
Endpoint, limits, regions, pricing.
Common Object Storage issues.
Object Storage endpoints.
All API updates and changes.