All Products
Search
Document Center

ApsaraDB RDS:RDS PostgreSQL IO performance burst

Last Updated:Apr 29, 2025

If your business experiences significant fluctuations with frequent peak periods, you can enable the IO performance burst feature for premium performance disks. This allows your disk IOPS to exceed the maximum IOPS limit, providing higher IO capabilities during business peaks to meet sudden business demands.

Background

Premium performance disk is a new storage type supported by RDS PostgreSQL. While maintaining compatibility with all features of ESSD disks, premium performance disks support the IO performance burst feature, which provides higher IOPS performance when storage capacity remains unchanged but IO pressure is high.

Feature introduction

After enabling IO performance burst for premium performance disks, the maximum IOPS and throughput limits of the instance will increase. Compared to instances without IO performance burst enabled, the differences are as follows:

Note

When IO performance burst is enabled on the primary node, it is also enabled on the secondary node. Read-only instances do not automatically enable this feature. You need to enable it on the details page of the read-only instance.

Scenario

Maximum IOPS

Maximum throughput

IO performance burst not enabled

min{50000+increased IOPS value,maximum IOPS of instance type,IOPS corresponding to maximum IO bandwidth of instance type,1800+50×storage space+increased IOPS value}

min{350+increased IO bandwidth value,maximum IO bandwidth of instance type,120+0.5×storage space+increased IO bandwidth value} (Unit: MB/s)

IO performance burst enabled

min{1000000,maximum IOPS of instance type,IOPS corresponding to maximum IO bandwidth of instance type}

min{4000,maximum IO bandwidth of instance type} (Unit: MB/s)

Note
  • If the throughput of an RDS instance reaches the upper limit, the IOPS of the instance is affected. If the IOPS of an RDS instance reaches the upper limit, the throughput of the instance is affected.

  • The above formulas do not apply to disk instances with general-purpose instance types because general-purpose instance types share resources and cannot guarantee maximum IOPS and maximum IO bandwidth.

  • For the maximum IOPS and maximum IO bandwidth of instance types in the above formulas, see RDS PostgreSQL primary instance type list.

  • The IOPS corresponding to the maximum IO bandwidth of instance type in the above formulas is calculated as: maximum IO bandwidth of instance type(Gbit/s)×1024×1024÷8÷8, where the first 8 from left to right represents 1Byte=8bits, and the second 8 represents that each read/write operation in RDS PostgreSQL uses 8KB of data.

Scenarios

I/O loads increase due to heavy workloads, large transactions, full table scans, and queries that return a large amount of data. In some cases, the I/O usage may approach or reach 100% even if the CPU, memory, and storage resources are sufficient. In these scenarios, you must upgrade the instance specifications or storage type of your RDS instance. For example, you can upgrade the storage type from PL1 ESSD to PL2 ESSD.

  • If you do not upgrade the instance specifications or storage type, the I/O loads reach the upper limit. As a result, the instance performance decreases, queries time out, and connection errors occur.

  • If you upgrade instance specifications, the I/O performance can meet your business requirements, but the CPU, memory, and storage resources are idle and wasted.

  • If you upgrade the storage type, the I/O usage is increased within a short period of time, but storage resources are wasted and costs are increased when the I/O loads decrease.

The IO performance burst feature of premium performance disks can solve these problems. After enabling this feature, when IO load is high, IO performance burst is automatically triggered to increase the IO limit. When IO load decreases, the IO limit automatically returns to normal, achieving serverless IO performance and avoiding waste of IO performance and cost.

Feature benefits

Compared to upgrading instance specifications or storage levels, the IO performance burst feature of premium performance disks offers multiple advantages and can significantly reduce costs.

  • Charges are only applied to the IO performance burst volume.

  • IO performance burst volume is calculated by second.

  • Free quota is provided for IO performance burst volume, and only the volume exceeding the free quota will be charged.

Billing

The premium performance disk IO performance burst feature is currently part of a fee reduction campaign. Starting from June 20, 2024, the IO performance burst feature can be used free of charge. For more information, see Fee reduction announcement. After the campaign ends, billing will begin, but you will still enjoy a partial free quota, with only the excess being charged.

Billing method

Premium performance disk fee = Storage fee + IO performance burst fee

  • Storage fee: fees for the storage capacity that you purchase for your RDS instance. The subscription and pay-as-you-go billing methods are supported. The billing method is the same as ESSD PL1 disks. For more information, see Billable items.

  • IO performance burst fee: Hourly IO performance burst fee for an instance = (Total IO performance burst volume of all instance nodes - Free quota) × IO performance burst unit price

    Note

    The IO performance burst volume fee for RDS instances (including primary and secondary nodes) only supports pay-as-you-go billing. It is charged when IO performance burst is enabled and the IO performance burst volume exceeds the free quota.

    • Sum of IO performance burst volumes of all instance nodes

      RDS edition

      Instance IO performance burst volume

      Basic Edition

      Single node IO performance burst volume

      High-availability Edition

      Primary node IO performance burst volume + Secondary node IO performance burst volume

      Cluster Edition

      Primary node IO performance burst volume + All secondary nodes IO performance burst volume

      Single node IO performance burst volume calculation method

      Scenario

      IO performance burst volume

      Example

      The IOPS exceeds the baseline but the throughput does not exceed the baseline.

      (IOPS - baseline IOPS) × duration

      Note

      For baseline IOPS, refer to the maximum IOPS value when IO performance burst is not enabled.

      If the instance IOPS exceeds the baseline IOPS by 4,000 and lasts for 2 seconds, the IO performance burst volume is 4,000 × 2 = 8,000.

      The throughput exceeds the baseline but the IOPS does not exceed the baseline.

      (throughput - baseline throughput) × 1024 × duration ÷ 16

      Note
      • For baseline throughput, refer to the maximum throughput value when IO performance burst is not enabled.

      • Throughput: the amount of data transferred per second. Unit: MB/s.

      16: the conversion coefficient.

      If the instance throughput exceeds the baseline throughput by 8 MB/s and lasts for 2 seconds, the IO performance burst volume is 8 × 1024 × 2 ÷ 16 = 1,024.

      Both the IOPS and throughput exceed their baselines.

      max{(IOPS - baseline IOPS) × duration, (throughput - baseline throughput) × 1024 × duration ÷ 16}

      Note

      Throughput: the amount of data transferred per second. Unit: MB/s.

      If (IOPS - baseline IOPS) × duration calculates to 8,000, and (throughput - baseline throughput) × 1024 × duration ÷ 16 calculates to 65,536, then the final IO performance burst volume is 65,536.

      Note
      • If the I/O loads on the primary RDS instance is heavy and an I/O burst is triggered, an I/O burst is also triggered on the secondary RDS instance to ensure data and service consistency. The IO performance burst volume of the secondary node is basically the same as that of the primary node.

      • The IO performance burst feature for read-only instances needs to be enabled separately, with the same billing standard as the primary node. If the read-only instance is of the High-availability Edition, the fee includes the IO performance burst fees for both the primary and secondary nodes.

    • Free quota:

      RDS provides a certain free usage quota for IO performance burst. When the used IO performance burst volume exceeds the free quota, RDS PostgreSQL charges a certain IO performance burst fee for different product editions.

      Product edition

      Free quota

      Basic Edition

      300,000 per hour

      High-availability Edition

      600,000 per hour

      Cluster Edition

      800,000 per hour

    • IO performance burst unit price: 0.0015 USD/10,000 IO

      Note

      If the IO performance burst volume exceeding the free quota is less than 10,000, it is billed as 10,000.

Billing example

Example scenario

An RDS instance that runs RDS High-availability Edition and provides a storage capacity of 1,000 GB resides in the China (Beijing) region. The baseline IOPS is 50,000 and the number of burstable I/O operations per second is 20,000. The burst lasts for 40 seconds every hour throughout a month.

The IO performance burst fee for this example scenario is calculated as follows:

IO performance burst volume

800,000 per hour

Free quota

600,000 per hour

Excess amount

200,000 per hour

Unit price

0.0015 USD/10,000 IO

Actual IO performance burst fee

0.0015 × (80-60) × 24 × 30 = 21.6 USD

The following table compares the fees of a general ESSD and a PL2 ESSD.

Note

The prices in the following table are provided for reference only. The actual prices are displayed in the ApsaraDB RDS console.

Storage type

Storage unit price (USD/month)

Storage fee (USD)

IO performance burst fee (USD)

Total fee (USD/month)

Premium performance disk

244.8

244.8 × 1 = 244.8

21.6

244.8 + 21.6 = 266.4

ESSD PL2 disk

489.6

489.6 × 1 = 489.6

N/A

489.6

Monthly savings of premium performance disk compared to ESSD PL2 disk: 489.6 - 266.4 = 223.2 USD

Note

Within a certain time period, the shorter the high IO duration of the instance, the more cost-effective the premium performance disk is compared to the ESSD PL2 disk.

Enable or disable IO performance burst

Note
  • Enabling or disabling the IO performance burst feature takes several minutes, depending on the instance usage (read and write traffic).

  • There is no transient connection during the process of enabling or disabling the IO performance burst feature, and it generally does not affect your business. In some cases, the IOPS of the RDS instance may fluctuate. We recommend that you perform this operation during off-peak hours.

  1. Go to the Instances page. In the top navigation bar, select the region in which the RDS instance resides. Then, find the RDS instance and click the ID of the instance.

  2. In the Basic Information section, click Premium Performance Disk Switch Settings after Storage Type. In the dialog box that appears, turn on or off the IO Performance Burst switch.

    image

References

  • For other storage types supported by RDS PostgreSQL, see Storage type introduction.

  • For more information about premium performance disks, see Premium performance disk.

  • When you need to address database disk IO performance bottlenecks, the Buffer Pool Extension (BPE) feature of premium performance disks can significantly improve database IO performance when facing large-scale or frequent data read/write requirements. For more information, see Buffer Pool Extension (BPE).

  • When you need to reduce storage costs, you can use the data archiving feature of premium performance disks to use Object Storage Service (OSS) as the storage medium for archiving cold data, significantly reducing storage costs. For more information, see Data archiving.

Related operations

API

Description

ModifyDBInstanceSpec

When configuring the IO performance burst feature for premium performance disks:

  • Make sure the DBInstanceStorageType parameter value is general_essd, which means the instance storage type is premium performance disk.

  • The BurstingEnabled parameter controls the enabling and disabling of the IO performance burst feature. Set it to true to enable and false to disable.

  • Do not modify other parameters, such as the instance type and storage capacity.