This topic describes how to use an accelerated endpoint of Object Storage Service (OSS) together with Function Compute and Serverless workflow to build a low-cost and highly efficient cross-border system of synchronized object transfer.


  • Cross-border transfer of object data is required.
  • The cost of cross-border data transfer must be strictly controlled.
  • Your applications are latency insensitive, and you can accept the synchronization latency caused by network jitters.
  • The system that is used to transfer data is elastic and can be scaled to handle large-scale file writes.

Solution overview

The following figure shows the architecture of the solution.

Low-cost cross-border data transfers - Figure

The following items describe the details:

  1. Elastic Compute Service (ECS) simulates applications to access OSS buckets in the China (Shanghai) and US (Silicon Valley) regions to produce and upload content.
  2. The content is uploaded to OSS buckets in China (Shanghai) by using an internal same-region endpoint.
  3. When an object is uploaded to an OSS bucket in China (Shanghai), Function Compute invokes a Serverless workflow. Each workflow uploads an object. This ensures that the object can be quickly synchronized by using the accelerated OSS endpoint in US (Silicon Valley).
  4. ECS simulates local users in US (Silicon Valley) to read OSS objects and check whether the data is correct.


  • Low O&M costs: Developers can focus on the code logic.
  • Low network cost: Compared with the combination of Express Connect and Cloud Enterprise Network (CEN), this solution is more cost-effective.
  • Low deployment cost of synchronization services: After an object is uploaded to an OSS bucket, Function Compute invokes a Serverless workflow. Serverless workflows are triggered on demand. You do not need to prepare ECS resources.
  • High elasticity and efficiency: One object triggers one Serverless workflow task. This ensures that resources are fully utilized and synchronization is performed in an efficient manner.