All Products
Search
Document Center

Tair:Overview

Last Updated:May 08, 2024

Tair provides multiple editions, series types, and architectures. This topic describes different types of instances and provides links to the relevant references.

Series type

Description

DRAM-based instance

DRAM-based instances use the multi-threading model. A DRAM-based instance provides approximately three times the read and write performance of a Community Edition instance that has the same specifications.

  • If the instance uses the standard architecture, the instance provides up to 64 GB of memory and supports up to approximately 300,000 queries per second (QPS).

  • If the instance uses the read/write splitting architecture, the instance provides up to 64 GB of memory and supports up to approximately 1,800,000 QPS.

  • If the instance uses the cluster architecture, the instance provides up to 16,384 GB (64 GB × 256 shards) of memory and the instance performance is calculated by using the following formula: Number of shards × Performance per shard. The performance of each shard is determined by the shard specifications.

DRAM-based instances support the local disk-based deployment mode.

  • Standard master-replica architecture: uses the multi-threading model and provides approximately three times the performance of a Community Edition instance that has the same specifications. Each instance provides up to 64 GB of memory and supports up to approximately 240,000 QPS.

  • Read/write splitting architecture: uses the multi-threading model, runs in the master-replica architecture, and consists of a master node and one or more read replicas. A read/writing splitting instance provides approximately three times the performance of a Community Edition instance that has the same specifications. Each instance provides up to 64 GB of memory and supports up to approximately 1,440,000 QPS.

  • Cluster architecture: uses the multi-threading model and provides approximately three times the performance of a Community Edition instance that has the same specifications. Each data shard uses the master-replica architecture. Each instance provides up to 4,096 GB (16 GB × 256 shards) of memory and supports up to approximately 61,440,000 QPS. This maximum QPS value is for reference only.

Persistent memory-optimized instance

Persistent memory-optimized instances do not use disks to implement data persistence. A persistent memory-optimized instance provides almost the same performance as a Community Edition instance in terms of throughput and latency while persisting each operation. A standard persistent memory-optimized instance can offer up to a 30% cost reduction compared with a Community Edition instance.

  • If the instance uses the standard architecture, the instance provides up to 64 GB of memory and supports up to approximately 100,000 QPS.

  • If the instance uses the read/write splitting architecture, the instance provides up to 64 GB of memory and supports up to approximately 600,000 QPS.

  • If the instance uses the cluster architecture, the instance provides up to 16,384 GB (64 GB × 256 shards) of memory and the instance performance is calculated by using the following formula: Number of shards × Performance per shard. The performance of each shard is determined by the shard specifications.

ESSD/SSD-based instance

ESSD/SSD-based instances store all data in disks and use memory to accelerate caching. ESSD/SSD-based instances reduce up to 85% of costs compared with ApsaraDB for Redis Community Edition instances, and deliver approximately 60% of the performance of ApsaraDB for Redis Community Edition instances. ESSD/SSD-based instances are suitable for warm and cold data storage scenarios that require compatibility with open source Redis, large capacity, and high access performance. Each instance provides up to 32,768 GB (128 GB × 256 shards) of memory and has a standard storage capacity of up to 327,680 GB (1,280 GB × 256 shards).

FAQ

  • Do I need to reserve memory for snapshots when I select specifications?

    No, you do not need to reserve memory for snapshots when you select specifications. Tair is sold at the instance level. The memory capacity of each specification type is the maximum memory that is available to you. The memory capacity includes the memory occupied by user data, the static memory consumed by your instance, and the memory occupied by network connections.

  • Each specification type has a maximum QPS value. What happens if the QPS of an instance exceeds the maximum value?

    This may cause an accumulation of requests. If the QPS reference value is exceeded for an extended period of time, we recommend that you choose higher specifications. For more information, see Change the configurations of an instance.

  • How do I check the specifications of an ApsaraDB for Redis instance by using InstanceClass?

    You can enter the value of InstanceClass in the search box in the upper part of an Alibaba Cloud document to search for the specifications.

  • How do I test the performance of Tair instances?

    You can test the performance of Tair instances by using the methods that are described in the performance whitepaper. For more information, see Performance whitepaper of Redis 5.0-compatible Tair DRAM-based instances.