When your RDS for MySQL instance hits IOPS limits during traffic spikes, I/O performance burst lets the disk exceed its provisioned baseline input/output operations per second (IOPS), giving you higher I/O capacity without permanently upgrading your instance. This is a pay-as-you-go feature for premium performance disks (Premium ESSDs) that charges only for the burst usage that exceeds a free hourly quota.
Prerequisites
Your RDS for MySQL instance must use premium performance disks (Premium ESSDs).
How it works
Enabling I/O burst raises the upper limits for both IOPS and throughput on premium performance disks:
| Scenario | Maximum IOPS | Maximum throughput (MB/s) |
|---|---|---|
| I/O burst disabled | min{50000, max IOPS of the instance type, 1800 + 50 × storage capacity} | min{350, max I/O bandwidth of the instance type, 120 + 0.5 × storage capacity} |
| I/O burst enabled | min{1,000,000, max IOPS of the instance type} | min{4000, max I/O bandwidth of the instance type} |
When throughput reaches its limit, IOPS is affected, and vice versa.
The formulas above do not apply to instances that use disks with general-purpose specifications, which use shared resources and cannot guarantee maximum IOPS or maximum I/O bandwidth.
For the maximum IOPS and I/O bandwidth values for your instance type, see RDS for MySQL Standard Edition (formerly X86) primary instance type list and RDS for MySQL Yitian Edition (formerly ARM) primary instance type list.
Primary and secondary node behavior
Enabling I/O burst on the primary node automatically enables it on the secondary node. Read-only instances are not affected — enable the feature separately on each read-only instance's product page.
Potential impacts
Enabling or disabling I/O burst does not cause transient connections and typically does not affect your services. In rare cases, IOPS may fluctuate. Perform this operation during off-peak hours.
The operation takes several minutes to complete. The exact duration depends on current read and write traffic.
Billing
The I/O performance burst feature is free of charge until September 16, 2025 (Singapore time). Starting from September 16, 2025, you will be charged for this feature. For more information, see the Official Billing Announcement.
Pricing
Unit price: USD 0.0015 per 10,000 I/O operations
I/O burst unit price: USD 0.0015 per 10,000 I/O operations.
Hourly fee formula:
Hourly I/O burst fee = (Total I/O burst usage of all nodes − Free quota) × Unit priceBilling is pay-as-you-go, calculated hourly. Usage is measured in units of 10,000 I/O operations. Partial units are rounded up.
Free quota
| Edition | Free quota |
|---|---|
| Basic Edition | 300,000 I/O operations per hour |
| High-availability Edition | 600,000 I/O operations per hour |
| Cluster Edition | 800,000 I/O operations per hour |
How instance I/O burst usage is calculated
The instance's total I/O burst usage is the sum of burst usage across all its nodes:
| Edition | Instance I/O burst usage |
|---|---|
| Basic Edition | I/O burst usage of the single node |
| High-availability Edition | I/O burst usage of the primary node + I/O burst usage of the secondary node |
| Cluster Edition | I/O burst usage of the primary node + I/O burst usage of all secondary nodes |
When the primary node experiences high I/O pressure and triggers a burst, the secondary node initiates a corresponding burst to maintain data and service consistency. The secondary node's I/O burst usage is nearly the same as the primary node's.
Read-only instances are billed separately under the same rules. If a read-only instance is a High-availability Edition instance, the fee includes burst costs for both its primary and secondary nodes.
Single node I/O burst usage:
Single node burst usage = (Portion exceeding baseline IOPS or throughput) × DurationBaseline I/O performance refers to the maximum IOPS and throughput when I/O burst is disabled.
Billing example
Scenario: A High-availability Edition instance with a 200 GB premium performance disk and a baseline IOPS of 12,000. Within one hour (00:00:00–01:00:00), bursts occur from 00:00:00 to 00:00:27 (27 seconds) and from 00:53:00 to 00:53:33 (33 seconds). During both bursts, IOPS reaches 22,000. At all other times, IOPS stays at or below the baseline.
| Item | Value |
|---|---|
| Total burst duration | 27s + 33s = 60s |
| I/O burst usage of a single node | (22,000 − 12,000) × 60s = 600,000 I/O operations |
| Total instance I/O burst usage | 600,000 (primary) + 600,000 (secondary) = 1,200,000 I/O operations |
Fee calculation for this instance (602,000 I/O operations in one hour):
| Item | Value |
|---|---|
| Actual I/O burst usage | 602,000 I/O operations |
| Free quota | 600,000 I/O operations |
| Usage exceeding the quota | 2,000 I/O operations (rounded up to 10,000) |
| Actual I/O burst fee | 10,000 × USD 0.0015 per 10,000 = USD 0.0015 |
Enable or disable I/O burst
Go to the RDS Instances page. In the top navigation bar, select the region where the instance resides, then click the instance ID.
In the Basic Information > Storage Type section, click Configure Premium ESSD next to Premium ESSD. Then enable or disable I/O Performance Burst.
API reference
Use the ModifyDBInstanceSpec operation to enable or disable I/O burst programmatically.
| Parameter | Description |
|---|---|
DBInstanceStorageType | Must be set to general_essd to confirm the instance uses premium performance disks |
BurstingEnabled | Set to true to enable or false to disable I/O burst |
All other parameters — including instance type and storage capacity — must match the current instance configuration and cannot be changed in the same call.