You can configure a lifecycle policy to automatically move infrequently accessed data to the IA storage class, reducing storage costs. When needed, you can access the data directly or retrieve it in bulk, enabling intelligent tiered storage for hot and cold data.
The lifecycle management feature is currently in invitational preview. To try it out, submit a ticket to request access.
How it works
CPFS lifecycle management provides automatic storage tiering based on file access time:
Automatic tiered storage: After you configure a policy, the system periodically scans the file system and automatically moves files that have not been accessed for a specified number of days from Standard storage to the IA storage class.
Transparent access mechanism: Data in the IA storage class can be accessed directly and supports all standard POSIX operations, so you do not need to change your access method.
Flexible retrieval policy: Based on your policy configuration, data can either be automatically moved back to Standard storage on first access or remain in the IA storage class and be read directly.
Configure a lifecycle policy
Step 1: Create a policy
You can create multiple policies for the same file system, each applying to different directories.
Log on to the CPFS console and navigate to the product page for your target file system.
In the navigation pane on the left, select Lifecycle. On the Lifecycle Policies tab, click Create Lifecycle Management Policy.
Configure the policy parameters:
Parameter
Description
Policy Name
Enter a custom policy name, such as “AI training data archiving policy.”
Scope
/ Entire file system: Applies to all files.
Enter directory path: Applies only to the specified directory, such as
/training/completed/.
Convert to IA storage class
Set a number of days (1–365). Files not accessed for longer than this period are automatically moved to IA storage class.
Unaccessed time = number of days since the file was last read or modified. Commands like ls and stat do not reset the timer.
Convert to Standard storage
Checked: Automatically moves the file back to Standard storage on access. Suitable for data that will be used frequently afterward.
Unchecked: Data is maintained in a state of infrequent access, suitable for infrequently accessed data.
Click OK to complete the creation.
Step 2: Verify policy activation
The policy takes effect during the next scan cycle after creation. The system scans once every 24 hours. The first scan time depends on when you created the policy.
Confirm that the policy status in the list is enabled.
After waiting 24 hours, check if IA storage capacity starts increasing under Basic Information on the file system product page.
On the Performance Monitoring page, view the success rate metric for conversions from Standard storage to IA storage.
Configuration reference
Based on different scenarios, we recommend the following configurations:
Scenario | Target | Convert to IA storage class | Convert to Standard storage | Description |
AI training data archiving |
| 30 days | Check. | Completed training datasets that may be used to reproduce experiments |
Autonomous driving data archiving |
| 60 days | Unchecked | Historical sensor data stored long-term and rarely accessed |
Log auditing archiving |
| 7 days | Unchecked | System logs requiring long-term retention and occasional viewing |
Full-file-system cold data archiving | / Entire file system | 90 days | Unchecked | Long-term archiving of the entire file system |
Accessing and Retrieving Low-Frequency Data
Data in the IA storage class is always available and supports two access methods:
Method 1: Direct access
You can read or write files in the IA storage class directly through the mount path. This method works well for temporary viewing or operations on a small number of files. It supports all standard file operations, such as read, write, modify, and delete, and provides an experience identical to accessing local files. Access performance depends on your IA storage policy configuration:
If you selected Convert to Standard storage: The first access incurs slight latency. The system asynchronously moves the file to Standard storage, and after conversion completes, you obtain full Standard storage performance.
If you did not select it: Every access reads data directly from the IA storage class, and the file remains in IA storage. Although performance is slightly lower than Standard storage, you continuously benefit from the lower cost of IA storage.
Method 2: Batch retrieval
When you need to access many IA storage files simultaneously—such as reading thousands of files for AI training—create a data retrieval task to batch convert files to Standard storage. This avoids cumulative latency that could affect business performance.
On the file system product page, navigate to Lifecycle > Data Retrieval Tasks.
Click Create Data Retrieval Task and configure the following parameters:
Task name: Enter a name for the task, such as "Training data prefetch - 2024Q1."
Target: Select / Entire file system or Enter directory path (must be an absolute path starting with "/").
Click OK to create the task.
You can check retrieval progress in the task list:
Pending: The task is created and waiting to run.
Running: Files are being converted from IA storage to Standard storage. You can view the progress percentage.
Completed: All files are successfully converted to Standard storage. Training can begin.
Partially failed: Some files failed to convert. You can view details and retry the failed files.
Failed: The task failed. Click Retry.
View storage usage
On the file system details page, you can view the following information:
Basic Information: The usage of Standard and IA storage classes.
Monitoring: Trends for capacity, IOPS, throughput, latency, and conversion success rate.
Quotas and limits
Item | Limit |
Number of lifecycle policies | Each Alibaba Cloud account can create up to 20 policies per region. |
Number of directories per policy | Up to 10 directories |
Number of data retrieval tasks | Each account can run up to 20 concurrent tasks per region. |
File system type | Only CPFS for Lingjun is supported. |
Billing
The IA storage class uses pay-as-you-go billing. It does not support subscription. There is no minimum storage duration. Deleting files incurs no extra fees.
Billing item | Charged? | Description |
Infrequent Access bucket | Billing | Billed hourly based on actual IA storage usage, settled monthly |
Standard to Infrequent Access | Free of charge | No traffic fees for moving data from Standard to IA storage |
Changing from Infrequent Access to Standard | Free of charge | No traffic fees for moving data from IA to Standard storage |
API requests | Free of charge | No charge for API requests to access files in IA storage |
Data retrieval tasks | No charge | No extra fees for batch retrieval tasks |
FAQ
Why hasn’t my file been moved to IA storage yet?
Lifecycle policies are not applied in real time. The system periodically scans the file system and applies policies. Execution typically begins within 24 hours after policy creation.
Will accessing files in IA storage cause latency?
If you selected Convert to Standard storage, the first access returns data with slight latency while the system asynchronously converts the storage type.
If you did not select this option, data is read directly from IA storage, with performance slightly lower than Standard storage.
How can I tell which files are in IA storage?
Currently, you cannot view a list of files in IA storage. You can check the total IA storage usage for your file system in the console.
What happens if I modify a file in IA storage?
When you modify a file in IA storage, the system automatically retrieves it to Standard storage before making changes. After modification, the file’s unaccessed timer resets.
Can I apply multiple policies to one directory?
Yes. If a file matches multiple rules, the system applies them in this priority order: Convert to IA storage > Convert to Standard storage.