ApsaraDB RDS supports four storage types: Enhanced SSDs (ESSDs), premium performance disks, high-performance local disks, and standard SSDs. Use this page to compare their characteristics and choose the right type for your workload.
Standard SSDs are being phased out. Use ESSDs for new instances.
Choose a storage type
| Storage type | Best for | Supported engines | I/O latency |
|---|---|---|---|
| ESSD (recommended) | General-purpose workloads requiring high reliability and fast scaling | MySQL, PostgreSQL, SQL Server, MariaDB | 100–200 µs |
| Premium performance disk | Workloads with burst traffic, large buffer pools, or cold data that can be archived | MySQL, PostgreSQL | 100–200 µs |
| High-performance local disk | I/O-intensive workloads requiring the lowest possible latency (flash sales, high-frequency trading) | MySQL only | 10–50 µs |
| Standard SSD (phased out) | — | MySQL, PostgreSQL, SQL Server, MariaDB | — |
Storage types
ESSD (recommended)
Enhanced SSDs (ESSDs) are built on a next-generation distributed block storage architecture using 25GE networks and Remote Direct Memory Access (RDMA) technology. They deliver up to 1 million random read/write IOPS and 100–200 µs I/O latency.
ESSDs are available at three performance levels (PLs):
PL1 ESSD: A PL1 performance level ESSD.
| Performance level | IOPS and throughput |
|---|---|
| PL1 ESSD | Baseline performance |
| PL2 ESSD | ~2× the IOPS and throughput of PL1 |
| PL3 ESSD | Up to 20× the IOPS and 11× the throughput of PL1; suited for workloads requiring extremely high concurrent I/O and stable read/write latency |
For detailed IOPS and throughput numbers by capacity tier, see ESSDs. For a cross-type performance comparison including maximum capacity, IOPS, and throughput, see Block storage performance.
Key characteristics:
Data durability: 99.9999999%, backed by multi-replica redundancy
Backup: ESSD snapshot backup; backup, read-only instance creation, and cloning complete in seconds
Scaling: Online scale-out in seconds with no downtime; scale-in supported for select MySQL and PostgreSQL instances
Editions: Supports Basic Edition (single-node) and High-availability Edition
Premium performance disk
Premium performance disks include all ESSD capabilities plus three additional features designed for demanding workloads:
[I/O performance burst](https://www.alibabacloud.com/help/en/rds/apsaradb-rds-for-mysql/i-o-performance-burst) (MySQL, PostgreSQL): Lets disk IOPS temporarily exceed its rated limit during traffic spikes, without requiring you to upgrade the disk tier permanently.
[Buffer Pool Extension (BPE)](https://www.alibabacloud.com/help/en/rds/apsaradb-rds-for-mysql/buffer-pool-extension-bpe) (MySQL, PostgreSQL): Expands the in-memory buffer pool onto disk, improving read performance for datasets that exceed available RAM.
[Data archiving](https://www.alibabacloud.com/help/en/rds/apsaradb-rds-for-mysql/rds-mysql-data-archiving-function) (MySQL, PostgreSQL): Moves infrequently accessed data to OSS, reducing storage costs by 80% compared to PL1 ESSDs. Suited for instances with tables that are rarely accessed or modified.
For more information, see Premium performance disks for RDS for MySQL. If you are using PostgreSQL, see Premium performance disks for RDS for PostgreSQL.
Key characteristics:
Data durability: 99.9999999%, backed by multi-replica redundancy
Backup: ESSD snapshot backup; backup, read-only instance creation, and cloning complete in seconds
Scaling: Online scale-out in seconds with no downtime; scale-in supported for select MySQL and PostgreSQL instances
Editions: Supports Basic Edition (single-node) and High-availability Edition
High-performance local disk
High-performance local disks are co-located on the same physical machine as the MySQL database engine, eliminating network round trips for I/O. This architecture delivers up to 150,000 IOPS, 10–50 µs I/O latency, and up to 16 TB of storage. Backup and recovery run 10× faster than self-managed databases.
Only High-availability Edition instances running MySQL support high-performance local disks.
Key characteristics:
Data durability: Hardware failures may cause data corruption; a secondary node is required for protection. The SLA for High-availability Edition instances can reach 99.995%.
Backup: XtraBackup physical backup; backup, read-only instance creation, and cloning time depends on disk size and may take hours
Scaling: Requires data copy; scale-out may take several hours and causes a transient connection interruption. For some instances that use high-performance local disks, disk space is coupled with the instance type and cannot be adjusted independently.
Editions: High-availability Edition only
Standard SSD (phased out)
Standard SSDs are distributed block storage devices that separate compute from storage. They are being phased out—end of sale for specific database engines took effect July 1, 2022. Use ESSDs for all new instances. For details, see the [EOS/Discontinuation] End of sale for the standard SSD storage type announcement.
Detailed comparison
Only High-availability Edition instances running MySQL support high-performance local disks.
| ESSD | Premium performance disk | High-performance local disk | |
|---|---|---|---|
| I/O performance | ★★★★★ IOPS determined by disk specs and instance type. I/O latency: 100–200 µs. | ★★★★★★ Extends ESSDs with I/O performance burst, BPE, and data archiving. IOPS determined by disk specs and instance type. I/O latency: 100–200 µs. | ★★★★★ IOPS determined by instance type. I/O latency: 10–50 µs. |
| Configuration flexibility | ★★★★★ Supports scale-out and scale-in. Note: scale-in is supported only for select MySQL and PostgreSQL instances. See Overview of instance configuration changes for MySQL instances or Change configurations for PostgreSQL instances. | ★★★★★ Supports scale-out and scale-in. Same conditions as ESSD apply. | ★★★★ Disk space can be adjusted independently on most instances. For some instances that use high-performance local disks, disk space is coupled with the instance type and cannot be adjusted separately. |
| Backup method | ESSD snapshot backup | ESSD snapshot backup | XtraBackup physical backup |
| Backup, read-only instance creation, and cloning | ★★★★★ Completes in seconds | ★★★★★ Completes in seconds | ★★★ Depends on disk size; may take hours |
| Scale-out duration | ★★★★★ Online upgrade; completes in seconds | ★★★★★ Online upgrade; completes in seconds | ★★★★ Requires data copy; may take hours |
| Impact of scaling out | None | None | A transient connection interruption occurs |
| Data durability | ★★★★★ 99.9999999%. Supports single-node Basic Edition instances. | ★★★★★ 99.9999999%. Supports single-node Basic Edition instances. | ★★★★ Hardware failures may cause data corruption. A secondary node is required for protection. The SLA for High-availability Edition instances can reach 99.995%. |
The reliability, durability, and read/write performance of all storage types comply with the product's Service-Level Agreement (SLA).
High-performance local disks: Use a primary/secondary High-availability Edition architecture. If the primary node fails, primary/secondary switchover completes within seconds.
Distributed disks (standard SSDs, ESSDs, premium performance disks): Ensure data reliability through multi-replica redundancy. High-availability Edition and Cluster Edition instances support automatic switchover within seconds.
View the storage type
View the storage type of an instance on its Basic Information page.
Purchase suggestions
We recommend that you prioritize ESSDs.
We recommend that you prioritize ESSDs.
Product support
For the full list of storage types and features supported by each engine, see: