This topic compares the online transaction processing (OLTP) performance between Archive Database (X-Engine engine) and Cluster (InnoDB engine) in PolarDB for MySQL 8.0.

PolarDB for MySQL Archive Database uses the storage engine X-Engine that is developed by Alibaba Cloud. X-Engine provides high data compression capabilities. This allows you to store a large amount of archived data at a low cost. Based on the internal best practices of Archive Database, when you store the same amount of data, Archive Database consumes only one-fifth to one-third of the storage of the InnoDB engine. For more information, see Archive database.

  • A cluster of Archive Database contains one primary node. By default, a cluster of Cluster contains one primary node and one read-only node. Therefore, to ensure comparability, this test is performed on only the primary node of the Cluster cluster.
  • Archive Database and Cluster use different engines. To make sure that both engines can achieve the highest performance, different numbers of threads are used to test nodes with the same specifications.
  • For more information about the test, see OLTP performance test.
The test results show the following information:
  • If the read-only nodes are the same, the read-only performance of Archive Database is almost the same as that of InnoDB.
  • If the node specifications are the same, the write performance of the Archive Database is 20% higher than that of the InnoDB database.

Comparison of read-only performance

read only
Note In the preceding test, the range_selects parameter is set to 0 and queries on all the ranges are disabled. Therefore, the performance is high.

Comparison of write performance

  • Comparison of write QPS QPS
  • Comparison of write TPS TPS
Note In the preceding queries:
  • Queries per second (QPS): the number of SQL statements that are executed per second in the database, including the BEGIN, COMMIT, INSERT, UPDATE, and DELETE statements.
  • Transactions per second (TPS): the number of transactions that are performed per second in the database. Only the successful committed transactions are counted.

For more information about how to test the write QPS and write TPS, see OLTP performance tests.

Comparison of read/write performance

  • Comparison of read/write QPSQPS
  • Comparison of read/write TPSTPS