AWS Database Blog
Amazon RDS Multi-AZ DB Clusters support new storage configurations
Amazon Relational Database Service (Amazon RDS) Multi-AZ DB Clusters are a great way to improve the performance and reduce the cost of demanding database workloads. Since RDS Multi-AZ DB Clusters were launched in March 2022, they have become an increasingly popular option for workloads that need more performance than a single RDS instance can support. However, RDS Multi-AZ DB Clusters have until recently required Amazon RDS Provisioned IOPS (io1) volumes.
On February 27, 2024, we announced support for Amazon RDS General Purpose gp3 volumes, and on March 6, 2024, we announced support for Amazon RDS Provisioned IOPs io2 Block Express volumes for RDS Multi-AZ DB Clusters. These new options allow you to deploy RDS Multi-AZ DB Clusters in several new configurations.
Amazon RDS gp3 volumes
Amazon RDS gp3 volumes were launched in 2022 with support for configuring IOPS independently from volume size, which gp2 volumes did not support. Additionally, gp3 volumes do not have the I/O burst credit bucket that gp2 volumes have, which means that they perform consistently even under prolonged periods of heavy load. This makes gp3 volumes suitable for many applications for which gp2 volumes were not.
Amazon RDS io2 Block Express volumes
Amazon RDS io2 Block Express volume allow you to deploy databases with a higher level of storage durability (99.999%, compared to 99.9% for io1 and gp3 volumes) and consistent latencies in the sub-millisecond range (compared to single-digit millisecond ranges for io1 and gp3 volumes).
Comparing RDS io1, RDS io2, and RDS gp3 volumes
The following table provides a high-level comparison of RDS io1, RDS io2, and RDS gp3 volumes.
| Volume Type | io2 | io1 | gp3 |
| Description | Highest performance within the RDS storage portfolio (IOPS, throughput, latency) Designed for latency-sensitive, transactional workloads |
Consistent storage performance (IOPS, throughput, latency) Designed for latency-sensitive, transactional workloads |
Flexibility in provisioning storage, IOPS, and throughput independently Balances price/performance for a wide variety of transactional workloads |
| Use cases | Business-critical transactional workloads that require sub-millisecond latency and sustained IOPS performance up to 256,000 IOPS | Transactional workloads that require sustained IOPS performance up to 256,000 IOPS | Broad range of workloads running on medium-sized relational databases in development and test environments |
| Latency | Sub-millisecond | Single-digit millisecond, provided consistently 99.9% of the time | Single-digit millisecond, provided consistently 99% of the time |
| Volume size | 100–65,536 GiB (16,384 GiB on RDS for SQL Server) | 100 GiB–64 TiB (16 TiB on RDS for SQL Server) | 20 GiB–64 TiB (16 TiB on RDS for SQL Server) |
| Maximum IOPS | 256,000 (64,000 on RDS for SQL Server) | 256,000 (64,000 on RDS for SQL Server) | 64,000 (16,000 on RDS for SQL Server) |
For more detailed information, refer to Amazon RDS DB instance storage.
All three volume types provide low latency and high throughput. However, RDS databases with gp3 volumes can have a maximum of 64,000 provisioned IOPS, whereas RDS databases with io1 or io2 volumes can have a maximum of 256,000 provisioned IOPS. As such, workloads that need more than 64,000 provisioned IOPS cannot be deployed on RDS databases with gp3 volumes. Additionally, io1 and io2 volumes provide more consistent latency, with io2 latencies in the sub-millisecond range, whereas gp3 volumes can experience periods of higher latency.
With this in mind, we recommend io2 volumes for workloads that are very sensitive to latency. If the AWS Region your database is deployed in does not yet support io2 volumes, io1 volumes are your next best choice. Finally, gp3 volumes are excellent choices for workloads where fewer IOPS are required, and small variances in latency are acceptable.
Summary
In this post, we discussed the availability of gp3 and io2 volumes for Amazon RDS Multi-AZ DB Cluster. These new options allow you to run RDS Multi-AZ DB Clusters in new configurations that may be suitable for your workloads, and may reduce your cost and increase your database performance.
If you have any feedback or questions, leave them in the comments section.
About the Author
Tim Gustafson is a Principal Database Solutions Architect at AWS, focusing on the open-source database engines and Amazon Aurora. When not helping customers with their databases on AWS, he enjoys spending time developing his own projects on AWS.