The tables below list the shard specifications for Tair (Enterprise Edition) memory-optimized (cloud-native edition) instances. Use these tables to select a shard size based on memory capacity, connections, bandwidth, and reference queries per second (QPS).
All specifications describe a single shard. For cluster and read/write splitting architecture instances, multiply the shard values by the number of shards to get the overall instance performance.
Instance types
The enterprise SSD (ESSD) listed in the specifications is used only for system operations—storing logs, temporary backup files, and Append-Only Files (AOFs). It is not used for data storage.
For the differences between direct connection mode and proxy mode, see Cluster architecture.
Cloud-native cluster architecture instances support adjusting the number of shard nodes. For more information, see Adjust the number of cluster shards.
Standard architecture and cluster architecture (direct connection mode)
The following specifications apply to standard architecture instances and cluster architecture instances in direct connection mode.
CPU coresEnterprise SSD (ESSD) (GB)| Specification name | Specification code (for API) | Memory (GB) | Bandwidth | Maximum connections | Reference QPS | ||
|---|---|---|---|---|---|---|---|
| 1 GB | tair.rdb.1g | 7 | 1 | 5 | 768 Mbps (96 MB/s) | 30,000 | 300,000 |
| 2 GB | tair.rdb.2g | 7 | 2 | 10 | 768 Mbps (96 MB/s) | 30,000 | 300,000 |
| 4 GB | tair.rdb.4g | 7 | 4 | 20 | 768 Mbps (96 MB/s) | 40,000 | 300,000 |
| 8 GB | tair.rdb.8g | 7 | 8 | 40 | 768 Mbps (96 MB/s) | 40,000 | 300,000 |
| 16 GB | tair.rdb.16g | 7 | 16 | 80 | 768 Mbps (96 MB/s) | 40,000 | 300,000 |
| 24 GB | tair.rdb.24g | 7 | 24 | 120 | 768 Mbps (96 MB/s) | 50,000 | 300,000 |
| 32 GB | tair.rdb.32g | 7 | 32 | 160 | 768 Mbps (96 MB/s) | 50,000 | 300,000 |
| 64 GB | tair.rdb.64g | 7 | 64 | 320 | 768 Mbps (96 MB/s) | 50,000 | 300,000 |
If read/write splitting is enabled on a standard architecture instance, use the specifications in the table below.
Cluster architecture (proxy mode) and read/write splitting architecture
The following specifications apply to cluster architecture instances in proxy mode and read/write splitting architecture instances.
CPU coresESSD (GB)| Specification name | Specification code (for API) | Memory (GB) | Bandwidth | Maximum connections | Reference QPS | ||
|---|---|---|---|---|---|---|---|
| 1 GB | tair.rdb.with.proxy.1g | 7 | 1 | 5 | 768 Mbps (96 MB/s) | 30,000 | 300,000 |
| 2 GB | tair.rdb.with.proxy.2g | 7 | 2 | 10 | 768 Mbps (96 MB/s) | 30,000 | 300,000 |
| 4 GB | tair.rdb.with.proxy.4g | 7 | 4 | 20 | 768 Mbps (96 MB/s) | 40,000 | 300,000 |
| 8 GB | tair.rdb.with.proxy.8g | 7 | 8 | 40 | 768 Mbps (96 MB/s) | 40,000 | 300,000 |
| 16 GB | tair.rdb.with.proxy.16g | 7 | 16 | 80 | 768 Mbps (96 MB/s) | 40,000 | 300,000 |
| 24 GB | tair.rdb.with.proxy.24g | 7 | 24 | 120 | 768 Mbps (96 MB/s) | 50,000 | 300,000 |
| 32 GB | tair.rdb.with.proxy.32g | 7 | 32 | 160 | 768 Mbps (96 MB/s) | 50,000 | 300,000 |
| 64 GB | tair.rdb.with.proxy.64g | 7 | 64 | 320 | 768 Mbps (96 MB/s) | 50,000 | 300,000 |
Performance details
Overall instance performance depends on the architecture:
| Architecture | Overall performance |
|---|---|
| Standard architecture | Same as the shard specification. |
| Cluster architecture (direct connection mode), cluster architecture (proxy mode), and read/write splitting | Overall performance = shard performance × number of shards. For cluster architecture (proxy mode) and read/write splitting architecture instances, the maximum bandwidth is capped at 16,384 Mbps (2,048 MB/s) and the maximum connections are capped at 500,000, regardless of the number of shards. |
Example: A four-shard instance using tair.rdb.with.proxy.1g has the following performance per shard:
-
CPU cores: 4 working cores + 3 background cores
-
Bandwidth: 768 Mbps (96 MB/s)
-
Maximum connections: 30,000
The overall instance performance is:
-
CPU cores: 28
-
Bandwidth: 3,072 Mbps (384 MB/s)
-
Maximum connections: 120,000
CPU core details
Each shard has 7 CPU cores: 4 working cores and 3 background cores.
-
Working cores process user service requests.
-
Background cores handle background tasks, including persistence, expiration acceleration, App Performance Analytics, and logging.
Bandwidth details
-
Bandwidth values in the tables above are per shard. The total bandwidth for cluster and read/write splitting architecture instances is the sum of all shard bandwidths.
-
For cluster architecture (proxy mode) and read/write splitting architecture instances, bandwidth is capped at 16,384 Mbps (2,048 MB/s). Adding more shards does not increase bandwidth beyond this limit. Cluster architecture (direct connection mode) instances have no overall bandwidth cap.
-
Bandwidth is bidirectional. A shard with 768 Mbps bandwidth has 768 Mbps upstream and 768 Mbps downstream.
-
The bandwidth limit applies to shard nodes and is independent of the network connectivity type.
For more information, see Bandwidth FAQ.
Maximum connections details
-
Proxy mode: The instance-level maximum is 500,000 connections. Adding more shards or nodes does not increase this limit.
-
Direct connection mode: Each shard supports up to 30,000 connections. The instance-level maximum equals the number of shards × 30,000.