ApsaraDB RDS:Primary ApsaraDB RDS instance types

Instance family

The instance family defines whether the resources for an instance are dedicated or shared, which directly determines the instance performance.

IOPS

IOPS (Input/Output Operations Per Second) measures the ability of a storage system to handle read and write requests and is a key metric for random I/O performance. The actual IOPS of an instancce is determined by its instance type, storage type, and storage capacity. You can find the maximum IOPS for an instance type in the primary instance type lists for reference. However, the actual maximum IOPS of the instance type is calculated as follows:

• For instances using Premium Local SSDs: The actual maximum IOPS is equal to the maximum IOPS defined in the instance type.

• For instances using cloud disks: The actual maximum IOPS is determined by the instance type, storage capacity, and storage type. Use the formulas in the following table to calculate the actual maximum IOPS:

  Storage type		Formula for actual maximum IOPS (Storage capacity unit: GB)
  Premium ESSD	With I/O performance burst enabled	min{Maximum IOPS for the instance type, 1000000}
  	With I/O performance burst disabled	min{Maximum IOPS for the instance type, 1800 + 50 × Storage capacity, 50000}
  ESSD	PL3	min{Maximum IOPS for the instance type, 1800 + 50 × Storage capacity, 1000000}
  	PL2	min{Maximum IOPS for the instance type, 1800 + 50 × Storage capacity, 100000}
  	PL1	min{Maximum IOPS for the instance type, 1800 + 50 × Storage capacity, 50000}
  Standard SSD		min{Maximum IOPS for the instance type, 1800 + 30 × Storage capacity, 25000}

Example:

Calculate the actual maximum IOPS of an RDS primary instance with the following specifications:

• Instance type: mysql.x2.large.2c

• Storage: PL1 ESSD of 20 GB in size

Based on the instance's storage type, the min{Maximum IOPS for the instance type, 1800 + 50 x Storage capacity, 50000} formula applies. The following table shows how to calculate the three values required in the formula:

According to the formula, the actual maximum IOPS of the instance is 2800, which is the smallest among the three values.

I/O bandwidth (throughput)

I/O bandwidth, also known as throughput, measures the ability of a storage system to handle sequential read and write operations and is a key metric for sequential I/O performance. The actual I/O bandwidth of an instancce is determined by its instance type, storage type, and storage capacity. You can find the maximum I/O bandwidth for an instance type in the primary instance type lists for reference. However, the actual maximum I/O bandwidth of the instance type is calculated using the formulas in the following table:

Example:

Calculate the actual maximum I/O bandwidth of an RDS primary instance with the following specifications:

• Instance type: mysql.x2.large.2c

• Storage: PL3 ESSD of 5,000 GB in size

Based on the instance's storage type, the min{Maximum I/O bandwidth for the instance type, 120 + 0.5 × Storage capacity, 4000} formula applies. The following table shows how to calculate the three values required in the formula:

According to the formula, the actual maximum I/O bandwidth of the instance is 192, which is the smallest among the three values.

Relationship between IOPS and I/O bandwidth

IOPS and I/O bandwidth are not independent metrics but rather two interconnected facets of storage performance. Their relationship is defined by the size of the I/O operations being performed as follows:

I/O bandwidth (MB/s) = IOPS × I/O Size (KB) / 1024.

This relationship means that the performance of an RDS instance is typically limited by either its IOPS or I/O bandwidth ceiling, depending on the application's I/O profile:

• IOPS-bound workloads: Applications performing many small I/O operations (such as 4 KB reads/writes) will often hit the maximum IOPS limit of the instance. At this point, the usage of the instance's I/O bandwidth is still well below its maximum, making IOPS the performance bottleneck.

• Bandwidth-bound workloads: Conversely, applications performing fewer, large I/O operations (such as 256 KB reads/writes) will saturate the available I/O bandwidth of the instance first. In this scenario, the IOPS of the instance remains relatively low, making I/O bandwidth the performance bottleneck.