Before creating a Tair (Redis OSS-compatible) instance, consider factors like performance, price, use cases, and workload to balance cost and stability. This topic covers product types, disaster recovery solutions, architectures, and instance specifications to guide your choice.
Free trial
Eligible new users can try Alibaba Cloud products for free. For more information, see Free Trial.
Quick recommendations
Click a link in the 'Recommended specifications' column below. You will only need to set basic options, such as the region and network, to create an instance.
Priority | Recommended specifications | Reasons |
Performance-first | Tair (Enterprise Edition), DRAM series (Redis 6.0-compatible), high availability, cluster enabled | In addition to the features of the Redis Open-Source Edition, this option provides:
|
Balance between performance and cost | Redis Open-Source Edition 6.0, high availability, cluster enabled |
|
Cost-first | Redis Open-Source Edition 6.0, high availability, cluster disabled |
|
Selection guide
Select product type
Tair (Redis OSS-compatible) offers both Redis Open-Source Edition and Tair (Enterprise Edition), an enterprise-grade in-memory database developed from Alibaba's business practices. Tair (Enterprise Edition) considers three core dimensions: access latency, persistence requirements, and overall cost. It offers the DRAM series, NVM series, and ESSD/SSD series to provide higher performance, additional data structures, and flexible storage options for various business needs.
For the commands and parameters supported by each series, see Command overview and Configure instance parameters.
The following table compares the product types.
Comparison item | Redis Open-Source Edition | Tair (Enterprise Edition) | ||
cloud-native (recommended) Classic | cloud-native | |||
Storage medium | Memory | |||
Compatible Redis versions | 5.0, 6.0, and 7.0 | 5.0, 6.0, and 7.0 | 6.0 | 6.0 |
Performance | 100% (benchmark) | 300% | 90% | Up to 60% |
Features | A managed Redis Open-Source Edition service. |
|
| Persists data to disk and uses memory to accelerate requests. |
Scenarios | General-purpose Redis use cases. |
| Scenarios that handle large volumes of data and require high cost-effectiveness and data reliability, such as the Internet of Things (IoT). | Ideal for storing large volumes of warm or cold data where cost is the top priority but high access performance is still required. Examples include file indexes and long-term message archives. |
For a detailed comparison of features and performance between Redis Open-Source Edition and Tair (Enterprise Edition), see Comparison between Tair (Enterprise Edition) and Redis Open-Source Edition.
Choose a disaster recovery solution
Tair (Redis OSS-compatible) offers single-zone, intra-city, and cross-region disaster recovery options. Choose the one that best suits your business requirements.
Disaster recovery solution | Description | Actions |
The master and replica nodes are deployed on different physical machines in the same zone for machine-level fault tolerance. | On the buy page, set Deployment Type to Single-zone Deployment. | |
The master and replica nodes are deployed in different zones (data centers) within the same region for data-center-level fault tolerance. | On the buy page, set Deployment Type to Dual-zone Deployment. | |
Multiple child instances are deployed in different regions to form a Global Distributed Cache instance. This solution provides region-level fault tolerance against events like natural disasters. For more information, see Global Distributed Cache. | For more information, see Create a distributed instance. |
Choose an architecture type
Tair (Redis OSS-compatible) supports two architecture types: standard architecture (cluster mode disabled) and cluster architecture, with an optional read/write splitting feature. Select an architecture based on your requirements for data volume, performance, and read/write capabilities.
By default, the following table describes instance architectures where shards are of the High-availability (Dual-replica) type. The standard architecture and cluster architecture also support the Single-node (Single-replica) type. However, the single-node type lacks data redundancy and is suitable only for testing or pure caching scenarios.
Instance architecture | Architecture model | Data distribution | Scenario | Read/write splitting |
Consists of a single shard using a master-replica replication model. | All data is stored in one shard. |
| This feature is optional. You can customize the number of read replicas. Each instance supports up to one master node and nine read replicas. | |
Consists of proxy nodes and multiple shards. Each shard uses a master-replica replication model. | Data is distributed across all shards. |
| This feature is optional. You can customize the number of read replicas. Each shard supports up to one master node and four read replicas. |
Instance specification
Select an appropriate instance specification and quantity (shard specification and number of shards) based on your estimated business metrics, such as capacity, bandwidth, number of connections, and queries per second (QPS). Recommended utilization: (Estimated usage ÷ Purchased capacity) < 80%.
When estimating capacity, you do not need to include memory overhead from copy-on-write operations or from advanced features like IP address whitelists, auditing, large key analysis, and hotkey detection. Alibaba Cloud covers this overhead, which does not count against your purchased instance capacity.
Large keys are a common issue in Redis. If a cluster has a large total capacity but small individual shards, a large key can exhaust the capacity of its shard.
The following table provides recommendations for selecting shard specifications for a cluster architecture.
Total capacity | Recommended shard specifications |
16 GB to 64 GB | 2 GB or higher |
64 GB to 256 GB | 4 GB or higher |
Greater than 256 GB | 8 GB or higher |
If your business needs change after purchase and the current specification becomes insufficient, you can change instance configuration at any time.