Updated on 2025-06-05 GMT+08:00

Process

This section describes how to use OBS+SFS Turbo for storage acceleration in AI scenarios. Figure 1 shows the process.

Figure 1 Process of using OBS+SFS Turbo for storage acceleration in AI scenarios
Table 1 Process description

No.

Step

Description

1

Plan networks and resources

Submit a service ticket to contact technical support engineers to support the planning.

2

Create resources

  1. Create a VPC and a subnet.
  2. Create an SFS Turbo HPC file system. Select a type based on the capacity and performance requirements. In AI scenarios, 250 MB/s/TiB or higher is recommended.
  3. Create an OBS bucket, with the storage class set to Standard and bucket policy set to Private.
  4. Create a ModelArts dedicated resource pool.

3

Complete basic configuration

  1. Configure network passthrough between ModelArts and SFS Turbo.
    1. Create an agency to authorize ModelArts to use SFS Turbo.
    2. Associate the ModelArts network with SFS Turbo.
  2. Configure SFS Turbo and OBS interworking.
  3. Configure auto data export from SFS Turbo to OBS.
  4. Configure the SFS Turbo data eviction policy.

4

Perform training

  1. Upload data to OBS and preload the data to SFS Turbo.
  2. Create training jobs.

5

Perform routine O&M

You can take the following O&M measures to ensure that the system can run efficiently.

  • Monitor SFS Turbo file system capacity through monitoring and alarms.
  • Monitor SFS Turbo file system performance.
  • Adjust the data eviction policy of the SFS Turbo file system.
  • Expand the capacity and improve the performance of the SFS Turbo file system.
  • Monitor the performance of the OBS bucket.

To know more about this solution or if you have any questions when using this solution, seek support through Solution Consultation.