Cloud-native instances are built on a new-generation control architecture that offers greater flexibility, faster scaling, and more advanced feature support than classic instances. Use this page to compare the two deployment modes and choose the one that fits your workload.
Classic instances support only Redis Open-Source Edition 5.0 and Tair Memory-optimized 5.0. If your workload requires Redis 6.0, Redis 7.0, or any Tair persistent memory or disk-based instance type, choose cloud-native.
Which should I choose?
Choose cloud-native if you need to:
-
Scale out quickly with minimal service disruption — cloud-native instances can scale without data migration when local resources are available.
-
Adjust shard count freely between 2 and 256 for cluster instances, rather than being constrained to 2, 4, or 8 fixed options.
-
Add or reduce replica nodes on demand (up to 9 secondary nodes for standard architecture, up to 4 per shard for cluster architecture).
-
Use proximity-based routing in multi-zone deployments to reduce read latency.
-
Keep failovers within the primary zone to avoid cross-zone latency spikes.
-
Use semi-synchronous replication, custom backup policies, or the ability to delete manual backups.
Classic instances are an older deployment mode with fixed shard configurations and limited scaling flexibility. They remain available but receive fewer feature updates. If you have an existing classic instance and want to migrate, see FAQ.
Architecture comparison
| Feature | Cloud-native instance | Classic instance |
|---|---|---|
| Architecture | Built on a new-generation control architecture for greater flexibility and scalability | Uses a traditional control architecture with fixed configurations |
| [Standard architecture](https://www.alibabacloud.com/help/en/redis/product-overview/standard-master-replica-instances) | Up to 1 primary node and 9 secondary nodes | 1 primary node and 1 secondary node only |
| [Cluster architecture](https://www.alibabacloud.com/help/en/redis/product-overview/cluster-master-replica-instances) | Shards: 2–256 (adjustable in single-shard increments); each shard supports up to 1 primary + 4 secondary nodes; no -ASK or -TRYAGAIN errors during scaling |
Shards: fixed at 2, 4, or 8; transient connections occur during scale-out |
| [Read/write splitting](https://www.alibabacloud.com/help/en/redis/product-overview/read-or-write-splitting-instances-1) | 1–9 adjustable read-only nodes; proximity-based access in multi-zone deployments; cluster (proxy mode) supports up to 4 read-only nodes per shard | Fixed at 1, 3, or 5 read-only nodes (architecture discontinued) |
| Scale-out | Fast scale-out without migration when local resources are sufficient; minimal service impact | Time-consuming |
| Disaster recovery | Multi-zone deployments support high-availability (HA) failover within the primary zone, preventing cross-zone latency increases | Multi-zone instances fail over to the secondary zone only; must manually switch back to the primary zone |
Feature support comparison
The tables below use the following symbols:
-
✔️ — Supported
-
❌ — Not supported
-
➖ — Not applicable to this architecture
Features available only in cloud-native instances
| Category | Feature | Cloud-native standard | Cloud-native cluster | Cloud-native read/write splitting | Classic standard | Classic cluster | Classic read/write splitting |
|---|---|---|---|---|---|---|---|
| Lifecycle management | Fast scale-out | ✔️ | ✔️ | ✔️ | ❌ | ❌ | ❌ |
| Seamless cluster scaling | ➖ | ✔️ | ➖ | ➖ | ❌ | ➖ | |
| Custom number of shards | ➖ | ✔️ | ➖ | ➖ | ❌ | ➖ | |
| Custom number of nodes | ✔️ | ✔️ | ✔️ | ❌ | ❌ | ❌ | |
| Custom number of read-only nodes for read/write splitting | ➖ | ➖ | ✔️ | ➖ | ➖ | ❌ | |
| Proximity-based access in multi-zone deployments | ➖ | ➖ | ✔️ | ➖ | ➖ | ❌ | |
| Backup and recovery | Custom backup policies for manual backups | ✔️ | ✔️ | ✔️ | ❌ | ❌ | ❌ |
| Delete manual backups | ✔️ | ✔️ | ✔️ | ❌ | ❌ | ❌ | |
| Extensions | Semi-synchronous mode | ✔️ | ✔️ | ✔️ | ❌ | ❌ | ❌ |
All features
Supported instance types
| Deployment mode | Instance type | Supported engine versions | Supported architectures |
|---|---|---|---|
| Cloud-native — Step 1: Create an instance | Redis Open-Source Edition | 7.0, 6.0, 5.0 | Standard, cluster, read/write splitting |
| Tair (Enterprise Edition) Memory-optimized | Compatible with Redis 7.0, 6.0, 5.0 | Standard, cluster, read/write splitting | |
| Tair (Enterprise Edition) Persistent Memory-optimized | Compatible with Redis 6.0 | Standard, cluster, read/write splitting | |
| Tair (Enterprise Edition) Disk-based | Compatible with Redis 6.0 | Standard, cluster | |
| Classic — Step 1: Create an instance | Redis Open-Source Edition | 5.0 | Standard, cluster, read/write splitting |
| Tair (Enterprise Edition) Memory-optimized | 5.0 | Standard, cluster, read/write splitting |
FAQ
How do I convert a classic instance to a cloud-native instance?
On the instance details page in the console, click Convert To Cloud-native. For step-by-step instructions, see Convert to the cloud-native deployment mode.