Docs

Torch Object Storage

Data subscription

Overview

The data subscription feature enables automated, policy-driven data replication. It allows you to securely and automatically synchronize AI-generated content (AIGC) from trusted services, such as Volcano Ark (VeArk), to your TOS buckets in near real time, without any development effort.

Use cases

Use this feature if you need to:

Replicate AI-generated content from VeArk: If you use VeArk, you can copy VeArk-generated images, videos, and 3D models to your TOS bucket for secondary processing, analysis, distribution, or archiving.

Features

The data subscription feature offers a variety of enterprise-grade features to address different data acquisition needs:

Policy-driven mechanism: It provides a visual console interface that allows you to filter the data to replicate by setting simple subscription rules, without writing any code.

Near real-time replication: It uses an asynchronous replication mechanism to make data accessible in your destination bucket within minutes of being written to the source. This meets the timeliness requirements of most business scenarios.

Reliability and retry: It offers highly reliable data replication. It also includes an automatic retry mechanism that handles temporary issues, such as network jitter.

Security and isolation: It implements secure, cross-account resource access based on Volcano Engine Identity and Access Management (IAM). This ensures data isolation among tenants and enforces the principle of least privilege.

Core concepts

Before you begin, understanding these core concepts will help you configure and manage data subscription rules effectively.

Concept	Description
Inference endpoint	The data subscription feature filters data at the inference endpoint level. In your rule, you must specify one or more inference endpoints.
Destination bucket and region	The TOS bucket in your account that is used to store VeArk inference outputs, and the region where the bucket resides.
Storage type	The TOS storage type used when the outputs are stored in your destination bucket, such as Standard Storage, Infrequent Access Storage, and Archive. You can specify it in the rule to optimize cost and performance.

Billing notes

You are responsible for all downstream costs associated with data subscription, including:

Item	Description
Storage cost	The cost of storing the generated content in your destination bucket. For more information, see Storage.
Request cost	The cost of making PutObject requests to write the generated content to your destination bucket. This charge is waived for the time being. For more information, see Requests.
Data transfer cost	The cost of transferring the generated content to your destination bucket. This charge is waived for the time being. For more information, see Data transfer.

FAQ

Q1: Is there an additional charge for the data subscription feature?

No, there are no additional fees for the feature. However, you will be charged for storing subscribed data in your bucket.

Q2: Will the subscribed file overwrite an existing file with the same name in the destination bucket?

It depends. When configuring a subscription rule, you can choose whether to overwrite or skip existing files with the same name. Design your object renaming rules carefully to avoid overwriting important data unintentionally.

Q3: Can I subscribe to data generated by services other than VeArk?

Not at this time. Currently, only outputs from VeArk are supported.

Q4: Can I synchronize historical data?

No. A data subscription rule only applies to data generated after the rule has been created.

Q5: What is the typical latency for data replication?

Typically, the P99 latency from the moment VeArk generates content to the moment it appears in your destination bucket is just a few minutes.

Last updated: 2026.08.20 17:53:16