Tair (Enterprise Edition) persistent memory-optimized instances use Storage Class Memory (SCM) to extend capacity beyond DRAM limits while keeping frequently accessed data in memory.
This page lists the instance types, per-shard specs, and performance scaling rules for persistent memory-optimized instances.
Instance types
The tables below list per-shard specifications. For cluster and read/write splitting instances, calculate overall performance using this formula:
Overall performance = Performance per shard x Number of shards
Enterprise SSDs (ESSDs) listed in these tables store system files only — logs, temporary file backups, and append-only files (AOFs). ESSDs are not used for data storage.
Shard specifications for standard architecture and cluster architecture in direct connection mode
ESSD (GB)| Specification name | Instance type code | CPU cores | Persistent memory (GB) | Bandwidth | Maximum connections | QPS (reference) | |
|---|---|---|---|---|---|---|---|
| 4 GB | tair.scm.standard.1m.4d | 3 | 4 | 12 | 768 Mbps (96 MB/s) | 20,000 | 100,000 |
| 8 GB | tair.scm.standard.2m.8d | 3 | 8 | 24 | 768 Mbps (96 MB/s) | 20,000 | 100,000 |
| 16 GB | tair.scm.standard.4m.16d | 3 | 16 | 48 | 768 Mbps (96 MB/s) | 20,000 | 100,000 |
| 32 GB | tair.scm.standard.8m.32d | 3 | 32 | 96 | 768 Mbps (96 MB/s) | 30,000 | 100,000 |
| 64 GB | tair.scm.standard.16m.64d | 3 | 64 | 192 | 768 Mbps (96 MB/s) | 30,000 | 100,000 |
The 4 GB instance type is not supported for the cluster architecture in direct connection mode. If you enable read/write splitting on a standard architecture instance, use the proxy mode and read/write splitting specifications below.
Shard specifications for cluster architecture in proxy mode and read/write splitting architecture
ESSD (GB)| Specification name | Instance type code | CPU cores | Persistent memory (GB) | Bandwidth | Maximum connections | QPS (reference) | |
|---|---|---|---|---|---|---|---|
| 4 GB | tair.scm.with.proxy.standard.1m.4d | 3 | 8 | 24 | 768 Mbps (96 MB/s) | 20,000 | 100,000 |
| 8 GB | tair.scm.with.proxy.standard.2m.8d | 3 | 8 | 24 | 768 Mbps (96 MB/s) | 20,000 | 100,000 |
| 16 GB | tair.scm.with.proxy.standard.4m.16d | 3 | 16 | 48 | 768 Mbps (96 MB/s) | 20,000 | 100,000 |
| 32 GB | tair.scm.with.proxy.standard.8m.32d | 3 | 32 | 96 | 768 Mbps (96 MB/s) | 30,000 | 100,000 |
| 64 GB | tair.scm.with.proxy.standard.16m.64d | 3 | 64 | 192 | 768 Mbps (96 MB/s) | 30,000 | 100,000 |
The 4 GB instance type is not supported for the cluster architecture in proxy mode. For the differences between direct connection mode and proxy mode, see Cluster architecture.
For cloud-native mode cluster instances, you can adjust the number of shards after creation. For details, see Adjust the number of shards for a cluster instance.
Performance details
| Architecture | How overall performance is calculated |
|---|---|
| Standard architecture | Same as the per-shard values listed above. |
| Cluster architecture (direct connection mode) | Per shard x number of shards. No overall bandwidth cap. |
| Cluster architecture (proxy mode) | Per shard x number of shards. Maximum bandwidth capped at 16,384 Mbps (2,048 MB/s); maximum connections capped at 500,000. |
| Read/write splitting | Per shard x number of shards. Maximum bandwidth capped at 16,384 Mbps (2,048 MB/s). |
Example: A cluster instance with four shards, each of type tair.scm.with.proxy.standard.2m.8d, has the following per-shard specs:
CPU cores: 3
Bandwidth: 768 Mbps (96 MB/s)
Connections: 10,000
The total instance performance is:
CPU cores: 12
Bandwidth: 3,072 Mbps (384 MB/s)
Maximum connections: 40,000
CPU core details
One CPU core per shard (or per read-only node) is reserved for background tasks to maintain service stability. The remaining cores handle your workload.
Bandwidth details
The bandwidth values in the tables represent the maximum bandwidth per shard. Total instance bandwidth equals the combined bandwidth of all shards.
For cluster instances in proxy mode and read/write splitting instances, total bandwidth is capped at 16,384 Mbps (2,048 MB/s). Adding more shards does not increase bandwidth beyond this limit. Cluster instances in direct connection mode have no overall bandwidth cap.
Bandwidth applies equally to upstream and downstream traffic. For example, an instance with 768 Mbps (96 MB/s) total bandwidth has 768 Mbps available for both upload and download.
Bandwidth limits apply to the data transfer capacity within the distributed system and are generally independent of the network connection type used by your client.
For common bandwidth questions, see FAQ about bandwidth.
Maximum connections
| Mode | Maximum connections per shard | Total instance limit |
|---|---|---|
| Proxy mode | — | 500,000 (fixed, regardless of shard count) |
| Direct connection mode | 10,000 | Shards x 10,000 (no fixed cap) |