This topic describes how to check the performance metrics of a PolarDB-X 1.0 instance and upgrade configurations to resolve insufficient performance issues.
- RT: RT reflects the performance of a single Structured Query Language (SQL) statement. You can fix this type of performance issue by using multiple methods, such as SQL optimization method.
- QPS: You can resolve QPS bottleneck issues by upgrading the configurations of the PolarDB-X 1.0 instance. You can upgrade the configurations to increase the QPS when the low latency and the high QPS are required for database access business.
The performance of PolarDB-X 1.0 is affected by the node performance of the compute layer and the storage layer. The insufficient performance of a node at the compute or storage layer leads to an overall performance bottleneck. For information about how to view the node performance at the storage layer, see Monitor storage nodes.
Determine whether a performance bottleneck of the instance occurs
The QPS and CPU performance of a PolarDB-X 1.0 instance is in positive correlation. When a performance bottleneck occurs in PolarDB-X 1.0, the main impact is that the CPU utilization of the instance remains high. If the CPU utilization exceeds 90% or remains higher than 80% for a consecutive period, this indicates that a performance bottleneck occurs in the current instance.
For information about how to view the performance metrics at the compute layer, see Monitor instances.
If no bottlenecks exist at the storage layer, you can determine that the current PolarDB-X 1.0 instance type cannot meet the QPS performance requirement of your business. In this case, you must upgrade the instance configurations.
Upgrade the configurations of PolarDB-X 1.0
QPS is an important metric for evaluating the PolarDB-X 1.0 instance type. Each instance type corresponds to a specific QPS reference value.
For more information about how to upgrade the configurations, see Change the instance configurations.