Tair (Redis® OSS-Compatible):Comparison between Tair（Redis OSS-compatible） and self-managed Redis

Item

Tair (Redis OSS-compatible)

Self-managed Redis

Security protection

Precautionary measures:

• Uses virtual private clouds (VPCs) to isolate the network.

• Uses whitelists for access control. For more information, see Step 2: Configure whitelists.

• Supports custom accounts and permissions. For more information, see Create and manage database accounts.

Precautionary measures:

• Requires a self-managed network security system, which is expensive and difficult to build.

• Comes with risks of data leaks due to security vulnerabilities in the default access configuration of Redis.

• Lacks an account authentication system.

In-process protection: supports TLS encryption.

In-process protection: requires a third-party tool to implement SSL encryption.

Post audit: supports audit logs. For more information, see Query audit logs.

Post audit: does not provide the auditing feature.

Backup and restoration

Tair (Enterprise Edition) DRAM-based instances support the data flashback feature that allows you to restore data to a specific point in time. For more information, see Use data flashback to restore data by point in time.

Only full data restoration is supported.

O&M management

• Supports more than ten groups of metrics and a minimum monitoring frequency of 5 seconds. For more information, see Metrics.

• Supports the configuration of alert rules. For more information, see Alert settings.

• Allows you to create instances of different architectures based on your business requirements and change the specifications or architecture of an existing instance.

• Supports large key analysis based on snapshots. This feature offers high accuracy and does not cause performance loss. For more information, see Use the offline key analysis feature.

• Requires a more complex third-party monitoring tool to monitor services.

• Stops providing services when you change the specifications or architecture of an instance. The operation to change the specifications or architecture of an instance is complex.

• Supports large key analysis based on sampling, which can be inaccurate.

Deployment and scaling

Supports elastic scaling and instant instance creation.

Requires a long period of time to complete hardware procurement, hosting of data centers, and machine deployment. You must manage node relationships when you add nodes.

High availability (HA)

• Provides the single-zone HA solution.

• Provides the zone-disaster recovery solution.

• Uses an independent central module to ensure HA. This module ensures stable and efficient decision-making and helps prevent split-brain issues.

• Allows you to deploy an HA architecture in Sentinel mode in a data center.

• Allows you to deploy a zone-disaster recovery architecture in Sentinel mode.

• Uses the Sentinel mode to ensure HA. The Sentinel mode is costly and has low decision-making efficiency during peak hours. Split-brain issues may occur, which can affect your services.

Kernel optimization

• Tair (Enterprise Edition) provides DRAM-based instances that use the multi-threading model. The performance of a DRAM-based instance is three times that of a standard instance with the same specifications.

• Tair (Enterprise Edition) provides ESSD/SSD-based and persistent memory-optimized instances to support large-capacity storage and command-level persistence.

• Instances of Redis 6.0 or later support multiple I/O threads to enhance performance. The performance may increase by up to twice, but CPU utilization also becomes high.

• Uses a storage service such as SSDB or Pika as persistent storage. However, these storage services are not fully compatible with the Redis protocol and allow the management of hot and cold data only at the key level. It is expensive to transfer large keys between memory and disks in these services.

Memory

Provides 100% memory availability. Memory overheads caused by disaster recovery, O&M management, scaling, and data persistence (such as write-time replication of forks) are borne by Alibaba Cloud and are not counted towards instance memory.

For example, if you create a 64 GB instance, the available memory capacity is 64 GB.

Reserves 25% to 40% of memory resources for disaster recovery, O&M management, and scaling.

For example, if you purchase two 64 GB Elastic Compute Service (ECS) instances to create Redis master/replica instances, the available total memory capacity is typically less than 45 GB.