> ## Documentation Index
> Fetch the complete documentation index at: https://docs.scitix.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# OSS Overview

> Learn about OSS use cases, Buckets, Access Method, and basic management capabilities.

OSS is an object storage service provided by GPU Cloud for storing and accessing object data. You can store datasets, model files, training outputs, log archives, and other unstructured data in OSS Buckets. Use the console to create and manage buckets. To access object data, use an SDK or command-line tool that supports the S3 protocol.

## Applicable Scenarios

OSS is suitable for storing data that does not rely on POSIX file system semantics and for reading and writing objects through the S3 protocol or command-line tools. Common scenarios include:

* Upload training datasets, model files, or inference input data.
* Download training outputs, model weights, or log archives.
* Access object data from a local environment, CES instance, or automated scripts.
* Archive unstructured data that does not need to be mounted as a local directory.

> [!TIP]
>
> If your workloads require multiple CES instances or workloads in a CKS cluster to share read and write access to the same data through local directories, the GPFS file system may be more suitable.

## Core Concepts

| Concept | Description |
| - | - |
| Buckets (Bucket) | Containers in OSS used to store objects. You can write objects to or read objects from a bucket only after creating it. Buckets are classified as high-performance or standard. When creating a bucket, only high-performance buckets can be created. Existing standard buckets can continue to be maintained, but new standard buckets cannot be created. |
| Object | A unit of data stored in OSS. It can be a file, a file object generated by copying a directory, or other unstructured data. |
| Access Keys | Authentication credentials for accessing OSS through the S3 protocol, including an Access Key and a Secret Key. |
| Endpoint | The service access address of OSS. Different regions use different endpoints. When configuring an SDK or command-line tool, use the endpoint corresponding to the region where the bucket is located. |

## Access Method

| Access Method | Applicable Tasks | Description |
| - | - | - |
| Console | Create Bucket, view bucket usage and object counts, and manage bucket settings. | Creating or managing buckets requires the relevant permissions. If you do not have permission, contact an administrator to obtain authorization. |
| S3 SDK or command-line tools | Upload or download objects, list objects, and upload or download directories. | Before use, obtain Access Keys, confirm the regional endpoint, and configure the relevant tools. |

## Usage workflow

**1. Create or Select a Bucket**

First, create a high-performance bucket in the console, or select an existing bucket. When creating a bucket, confirm parameters such as the region, Bucket Name, Versioning, and size.

When creating a bucket, only new high-performance buckets are supported. Existing standard buckets can continue to be maintained, but new standard buckets cannot be created. For more information, see [Create and manage OSS buckets](https://docs.scitix.ai/gpu-cloud/storage/create-oss-bucket.md).

**2. Configure access control, compliance retention, and cross-origin access rules**

Based on business access methods and data protection requirements, configure access control, compliance retention, or cross-origin access rules in the bucket settings as needed.

* To grant specified users access to a specific path prefix, add a directory policy on the **Access Control** tab.
* To prevent objects from being deleted before the end of the protection period, configure **Compliance Retention**.
* When a web application needs to access bucket resources across origins from a browser, configure cross-origin access rules on the **Overview** tab.

**3. Prepare Access Credentials and Tools**

Before accessing object data through the S3 SDK, ossctl, or other command-line tools, prepare Access Keys and configure the tool using the Endpoint corresponding to the Region where the Buckets are located.

If you use ossctl, first [Configure ossctl](https://docs.scitix.ai/gpu-cloud/storage/configure-ossctl.md).

**4. Upload, Download, or List Objects**

After configuring the tool, you can use ossctl or other tools that support the S3 protocol to upload objects, download objects, list objects, or upload and download directories.

For instructions on using ossctl to access objects, see [Upload and Download OSS Objects Using ossctl](https://docs.scitix.ai/gpu-cloud/storage/upload-download-oss-object.md).

**5. View usage and maintain data security**

During use, you can view bucket usage and the number of objects in the console, and adjust access permissions, cross-origin access rules, or retention policies as business needs change.

Before deleting an object or bucket, ensure that any data to be retained has been backed up or migrated.

## Security and Data Protection

* Store your Access Key and Secret Key securely. Do not include them in public code repositories, images, script templates, or logs.
* If an access key is compromised, disable or delete the key as soon as possible, and create a new access key.
* Deleting objects or Buckets may result in unrecoverable data loss. Before performing a delete or overwrite operation, ensure that any data to be retained has been backed up or migrated.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.