edit-icon download-icon

Data Migration Specifications

Last Updated: Dec 27, 2017

Note: This test report only serves as a performance reference for data migration. It is not the criteria of the product SLA.

Introduction

This performance test report provides a basic reference for the performance of different specifications of migration instances. You can use this report when selecting a specification for your migration instance.

Glossary

  • Specification

    DTS provides several levels of specification with different migration performance. The specification is based on the incremental data replication performance.

  • Table count

    The total number of tables in the test model.

  • Record size

    The size of each record.

  • RPS

    The amount of records being replicated per second. The records indicate data rows modified by SQL statements, such as insert, udpate and delete. Commit statement is also counted.

Test model

Procedure

  1. Create a migration task with incremental data replication between two RDS instances.

  2. Perform stress tests on the source RDS instance to check the performance of incremental data replication.

Environment

Instance RDS Configuration Reference Performance Limit
Source instance
  • Specification: rds.mys2.8xlarge
  • Memory (MB): 48,000
  • Max connections: 2,000
  • Max QPS: 18,000
  • Max IOPS: 14,000
Target instance
  • Specification: rds.mys2.8xlarge
  • Memory (MB): 48,000
  • Max connections: 2,000
  • Max QPS: 18,000
  • Max IOPS: 14,000

Conditions

  • The number of test tables is 20.

  • The record size is 1 KB.

  • All test tables have primary keys.

  • Each transaction contains two DML operations and one commit statement on average. The proportional amount between insert, delete, and update statements is 3:1:2.

Results

Note: The unit of the network latency is millisecond.

Source Instance Region > Target Instance Region Network Latency Specification TPS QPS
East China 1 (Hangzhou) > East China 1 (Hangzhou) 0.26 small 2,566 8,981
East China 1 (Hangzhou) > East China 1 (Hangzhou) 0.26 medium 4,726 16,541
East China 1 (Hangzhou) > East China 1 (Hangzhou) 0.26 large 6,378 23,204
East China 1 (Hangzhou) > North China 1 (Qingdao) 26 small 2,469 8,641
East China 1 (Hangzhou) > North China 1 (Qingdao) 26 medium 4,856 16,996
East China 1 (Hangzhou) > North China 1 (Qingdao) 26 large 5,439 20,400
East China 1 (Hangzhou) > North China 2 (Beijing) 26 small 2,533 8,866
East China 1 (Hangzhou) > North China 2 (Beijing) 26 medium 5,038 17,633
East China 1 (Hangzhou) > North China 2 (Beijing) 26 large 6,829 26,100
East China 1 (Hangzhou) > West US (Silicon Valley) 175 small 1,753 6,135
East China 1 (Hangzhou) > West US (Silicon Valley) 175 medium 2,837 9,929
East China 1 (Hangzhou) > West US (Silicon Valley) 175 large 3,884 15,500
Asia-Pacific (Singapore) > West US (Silicon Valley) 198 small 1,104 4,000
Asia-Pacific (Singapore) > West US (Silicon Valley) 198 medium 1,724 6,334
Asia-Pacific (Singapore) > West US (Silicon Valley) 198 large 2,256 8,300

The test result indicates the ultimate performance of different specifications. To obtain ultimate performance, the following conditions are required:

  • Each synchronization table under the synchronization instance has a primary key.
  • No update hotspot.
  • No performance bottlenecks between the source and target instances.
Thank you! We've received your feedback.