A DTS Serverless instance is a Data Transmission Service (DTS) instance that automatically scales resources based on your workload — so you pay only for what you use and avoid over-provisioning.
How it works
Serverless instances measure capacity in DTS Units (DUs). Each DU represents a bundle of compute, memory, and network resources. The system continuously monitors workload signals — records per second (RPS), CPU, memory utilization, and network — and adjusts the DU count in real time, typically within minutes. After each adjustment, the system waits 60 seconds before checking whether the new capacity meets the current load.
Billing follows the DU count: the service generates one bill per hour based on actual resource usage, not a fixed specification.
In workloads with fluctuating data volumes, a standard instance either wastes resources during quiet periods or falls short during spikes. A Serverless instance tracks the actual load throughout:
Advantages
High elasticity — Resource specifications adjust in real time (within minutes) as data transfer volumes change, improving resource utilization and reducing overall resource consumption.
Fully managed — No manual upgrades or downgrades are needed, which improves O&M efficiency and lowers O&M costs for O&M staff and developers.
Low cost — True pay-as-you-go billing with no fixed-resource pricing. Resources scale dynamically based on actual load, saving significant costs.
Use cases
Serverless instances are a good fit when:
Data transfer volume is high at the start of a project but levels off over time.
Traffic follows clear peak and off-peak patterns and you want to reduce the cost of continuous full-capacity running.
You have existing data synchronization tasks and want to convert standard sync instances to Serverless, or add new Serverless instances alongside them.
Transfer volume is unpredictable and hard to size in advance.
You are an individual developer.
Supported regions
Serverless instances are available in the following regions: China (Hangzhou), China (Shanghai), China (Qingdao), China (Beijing), China (Zhangjiakou), China (Ulanqab), China (Shenzhen), China (Guangzhou), China (Wuhan - Local Region), China (Chengdu), China (Hong Kong), Singapore, Malaysia (Kuala Lumpur), Indonesia (Jakarta), Japan (Tokyo), and Germany (Frankfurt).
Performance reference
The table below shows the RPS ceiling at each DU count. Actual throughput varies depending on network conditions, source and destination database performance, and latency.
| DU count | Performance ceiling (RPS) |
|---|---|
| 1 | 200 |
| 2 | 1,000 |
| 4 | 3,000 |
| 8 | 9,000 |
| 16 | 17,000 |
| 32 | 35,000 |
Limitations
Generally available
Only data synchronization tasks support Serverless instances. Data migration and change tracking tasks do not.
The minimum DU value must be one of: 1, 2, 4, 8, 16, or 32. The maximum DU value must be one of: 2, 4, 8, 16, or 32.
In preview (available to select users only)
The following configurations are undergoing grayscale testing and are not yet available to all users:
Cross-region or cross-border replication: Only Southwest 1 (Chengdu) to South China 3 (Guangzhou) and the reverse direction are supported. All other region pairs are not supported.
Two-way synchronization.
Source Database Type is Tair/Redis, PolarDB-X 1.0, PolarDB-X 2.0, or Sharded Cluster MongoDB.
Destination Database Type is Tair/Redis.
Billing
For pricing details, see Serverless instance billing method.
What's next
FAQ
Can I create Serverless instances for data migration or change tracking tasks?
No. Only data synchronization tasks support Serverless instances.