Instance types for dual-replica cluster architecture instances of ApsaraDB for Redis (Redis Open-Source Edition). Each shard consists of one primary node and one replica node.
Instance types
Instance types are available with the following numbers of shards: 2 shards, 4 shards, 8 shards, 16 shards, 32 shards, 64 shards, 128 shards, and 256 shards.
All specs scale linearly with shard count — total bandwidth caps at 16,384 Mbps (2,048 MB/s) and proxy-mode connections cap at 500,000 regardless of shard count.
2 shards
4 CPU cores total (2 per shard).
| Instance type name | Instance type code (for API calls) | Memory per shard (GB) | Proxy mode — max connections | Direct connection mode — max connections | Total bandwidth | Bandwidth per shard | Reference QPS |
|---|---|---|---|---|---|---|---|
| 4 GB Cluster Edition (2 shards) | redis.logic.sharding.2g.2db.0rodb.4proxy.default | 2 | 240,000 | 20,000 | 1,536 Mbps (192 MB/s) | 768 Mbps (96 MB/s) | 200,000 |
| 8 GB Cluster Edition (2 shards) | redis.logic.sharding.4g.2db.0rodb.4proxy.default | 4 | 240,000 | 20,000 | 1,536 Mbps (192 MB/s) | 768 Mbps (96 MB/s) | 200,000 |
| 16 GB Cluster Edition (2 shards) | redis.logic.sharding.8g.2db.0rodb.4proxy.default | 8 | 240,000 | 20,000 | 1,536 Mbps (192 MB/s) | 768 Mbps (96 MB/s) | 200,000 |
4 shards
8 CPU cores total (2 per shard).
| Instance type name | Instance type code (for API calls) | Memory per shard (GB) | Proxy mode — max connections | Direct connection mode — max connections | Total bandwidth | Bandwidth per shard | Reference QPS |
|---|---|---|---|---|---|---|---|
| 8 GB Cluster Edition (4 shards) | redis.logic.sharding.2g.4db.0rodb.4proxy.default | 2 | 480,000 | 40,000 | 3,072 Mbps (384 MB/s) | 768 Mbps (96 MB/s) | 400,000 |
| 16 GB Cluster Edition (4 shards) | redis.logic.sharding.4g.4db.0rodb.4proxy.default | 4 | 480,000 | 40,000 | 3,072 Mbps (384 MB/s) | 768 Mbps (96 MB/s) | 400,000 |
| 24 GB Cluster Edition (4 shards) | redis.logic.sharding.6g.4db.0rodb.4proxy.default | 6 | 480,000 | 40,000 | 3,072 Mbps (384 MB/s) | 768 Mbps (96 MB/s) | 400,000 |
| 32 GB Cluster Edition (4 shards) | redis.logic.sharding.8g.4db.0rodb.4proxy.default | 8 | 480,000 | 40,000 | 3,072 Mbps (384 MB/s) | 768 Mbps (96 MB/s) | 400,000 |
| 64 GB Cluster Edition (4 shards) | redis.logic.sharding.16g.4db.0rodb.4proxy.default | 16 | 480,000 | 40,000 | 3,072 Mbps (384 MB/s) | 768 Mbps (96 MB/s) | 400,000 |
| 128 GB Cluster Edition (4 shards) | redis.logic.sharding.32g.4db.0rodb.8proxy.default | 32 | 480,000 | 40,000 | 3,072 Mbps (384 MB/s) | 768 Mbps (96 MB/s) | 400,000 |
8 shards
16 CPU cores total (2 per shard).
| Instance type name | Instance type code (for API calls) | Memory per shard (GB) | Proxy mode — max connections | Direct connection mode — max connections | Total bandwidth | Bandwidth per shard | Reference QPS |
|---|---|---|---|---|---|---|---|
| 32 GB Cluster Edition (8 shards) | redis.logic.sharding.4g.8db.0rodb.8proxy.default | 4 | 500,000 | 80,000 | 6,144 Mbps (768 MB/s) | 768 Mbps (96 MB/s) | 800,000 |
| 48 GB Cluster Edition (8 shards) | redis.logic.sharding.6g.8db.0rodb.8proxy.default | 6 | 500,000 | 80,000 | 6,144 Mbps (768 MB/s) | 768 Mbps (96 MB/s) | 800,000 |
| 64 GB Cluster Edition (8 shards) | redis.logic.sharding.8g.8db.0rodb.8proxy.default | 8 | 500,000 | 80,000 | 6,144 Mbps (768 MB/s) | 768 Mbps (96 MB/s) | 800,000 |
| 128 GB Cluster Edition (8 shards) | redis.logic.sharding.16g.8db.0rodb.8proxy.default | 16 | 500,000 | 80,000 | 6,144 Mbps (768 MB/s) | 768 Mbps (96 MB/s) | 800,000 |
16 shards
32 CPU cores total (2 per shard).
| Instance type name | Instance type code (for API calls) | Memory per shard (GB) | Proxy mode — max connections | Direct connection mode — max connections | Total bandwidth | Bandwidth per shard | Reference QPS |
|---|---|---|---|---|---|---|---|
| 64 GB Cluster Edition (16 shards) | redis.logic.sharding.4g.16db.0rodb.16proxy.default | 4 | 500,000 | 160,000 | 12,288 Mbps (1,536 MB/s) | 768 Mbps (96 MB/s) | 1,600,000 |
| 96 GB Cluster Edition (16 shards) | redis.logic.sharding.6g.16db.0rodb.16proxy.default | 6 | 500,000 | 160,000 | 12,288 Mbps (1,536 MB/s) | 768 Mbps (96 MB/s) | 1,600,000 |
| 128 GB Cluster Edition (16 shards) | redis.logic.sharding.8g.16db.0rodb.16proxy.default | 8 | 500,000 | 160,000 | 12,288 Mbps (1,536 MB/s) | 768 Mbps (96 MB/s) | 1,600,000 |
32 shards
64 CPU cores total (2 per shard).
| Instance type name | Instance type code (for API calls) | Memory per shard (GB) | Proxy mode — max connections | Direct connection mode — max connections | Total bandwidth | Bandwidth per shard | Reference QPS |
|---|---|---|---|---|---|---|---|
| 128 GB Cluster Edition (32 shards) | redis.logic.sharding.4g.32db.0rodb.32proxy.default | 4 | 500,000 | 320,000 | 16,384 Mbps (2,048 MB/s) | 768 Mbps (96 MB/s) | 3,200,000 |
| 192 GB Cluster Edition (32 shards) | redis.logic.sharding.6g.32db.0rodb.32proxy.default | 6 | 500,000 | 320,000 | 16,384 Mbps (2,048 MB/s) | 768 Mbps (96 MB/s) | 3,200,000 |
| 256 GB Cluster Edition (32 shards) | redis.logic.sharding.8g.32db.0rodb.32proxy.default | 8 | 500,000 | 320,000 | 16,384 Mbps (2,048 MB/s) | 768 Mbps (96 MB/s) | 3,200,000 |
64 shards
128 CPU cores total (2 per shard).
| Instance type name | Instance type code (for API calls) | Memory per shard (GB) | Proxy mode — max connections | Direct connection mode — max connections | Total bandwidth | Bandwidth per shard | Reference QPS |
|---|---|---|---|---|---|---|---|
| 512 GB Cluster Edition (64 shards) | redis.logic.sharding.8g.64db.0rodb.64proxy.default | 8 | 500,000 | 640,000 | 16,384 Mbps (2,048 MB/s) | 768 Mbps (96 MB/s) | 6,400,000 |
128 shards
256 CPU cores total (2 per shard).
| Instance type name | Instance type code (for API calls) | Memory per shard (GB) | Proxy mode — max connections | Direct connection mode — max connections | Total bandwidth | Bandwidth per shard | Reference QPS |
|---|
256 shards
512 CPU cores total (2 per shard).
| Instance type name | Instance type code (for API calls) | Memory per shard (GB) | Proxy mode — max connections | Direct connection mode — max connections | Total bandwidth | Bandwidth per shard | Reference QPS |
|---|
Usage notes
CPU cores
One CPU core per shard or read replica is reserved for background tasks. This applies to both cluster instances and read/write splitting instances.
Bandwidth
Total bandwidth is the sum across all shards. Per-shard limits:
For Redis Open-Source Edition:
1 GB memory or less: 384 Mbps (48 MB/s)
More than 1 GB memory: 768 Mbps (96 MB/s)
For Tair: 768 Mbps (96 MB/s)
Bandwidth cap: Total bandwidth for a cluster instance using the default database proxy endpoint is capped at 16,384 Mbps (2,048 MB/s). Adding more shards does not increase it beyond this limit.
Upstream and downstream: The listed bandwidth applies equally to both directions. For example, a 384 Mbps (48 MB/s) instance has 384 Mbps available for both uploads and downloads.
Internal vs. public: The bandwidth listed is internal bandwidth. Use an internal connection to eliminate public network overhead and get the full rated bandwidth.
Adjusting bandwidth: If your instance hits traffic peaks, manually increase instance bandwidth as needed.
To handle sustained high traffic, enable direct connection mode. For details, see Enable direct connection mode. Direct connection mode is available for cluster instances only. After enabling it:
Total bandwidth: Calculated as (bandwidth per shard) x (number of shards). For example, a 128-shard instance where each shard has more than 1 GB of memory gets 128 x 96 MB/s = 12,288 MB/s.
Maximum connections: Calculated as (connections per shard) x (number of shards). Redis Open-Source Edition allows 10,000 connections per shard; Tair allows 30,000.
Maximum connections
Proxy mode Capped at 500,000 regardless of shard count.
Instances created before March 1, 2020 have a proxy-mode cap of 200,000.
Direct connection mode 10,000 connections per shard. Total = 10,000 x (number of shards).
FAQ
My instance's actual bandwidth differs from the spec listed here. Why?
Bandwidth limits for some instance types change after product upgrades. To get the updated bandwidth without extra cost, change the instance configuration and select your current instance type as the target — the system applies the new limits at no additional charge.
Write operations are failing on my cluster instance, but memory isn't full. Why?
Redis distributes keys across shards using a hash algorithm. Large keys can skew this distribution — one shard fills up while others stay empty, causing writes to that shard to fail.
Use the Performance monitoring feature to check memory usage per shard and identify hot shards. See How to view performance metrics such as memory usage of child nodes in a Redis cluster for details.
Does "8proxy" in an instance type code like `redis.logic.sharding.1g.8db.0rodb.8proxy.default` mean there are eight proxy nodes?
No. The 8proxy segment does not represent the number of proxy nodes. Proxy capacity and bandwidth are guaranteed to meet the specs listed in the tables — you don't need to track the actual node count.