All Products
Search
Document Center

Tair (Redis® OSS-Compatible):Disk-based instances

Last Updated:Mar 30, 2026

Tair ESSD/SSD-based instances store data on ESSDs (Enhanced SSDs) or local SSDs instead of volatile memory, delivering large-capacity, persistent database storage at up to 85% lower cost than Redis Open-Source Edition instances. These instances are compatible with most Redis 6.0 data structures and commands, making them a drop-in option for workloads with warm and cold data patterns.

Background

Redis stores all data in volatile memory. As your data volume grows, this creates three problems:

  • Cost: Rarely accessed data still occupies expensive memory, making in-memory storage less cost-effective over time.

  • Persistence: You need to integrate additional databases or storage systems to handle data durability.

  • Capacity: Standalone instances and cluster size limits cap how much data you can store.

ESSD/SSD-based instances address all three by using disk-based storage with the TairDB storage engine, which combines disks and memory to provide an optimal balance between data persistence and quick access to data.

Key features

Feature Details
Compatibility Compatible with most data structures and commands of Redis 6.0. See Limits on commands supported by Tair.
Cost Up to 85% lower cost compared with Redis Open-Source Edition instances.
Performance Approximately 60% of the throughput of Redis Open-Source Edition instances. See performance whitepapers for ESSD-based and SSD-based instances.
Capacity Storage capacity can reach hundreds of terabytes.
Data reliability Data is persisted to ESSDs or local SSDs, eliminating the risk of data loss from memory volatility.
Replication Supports semi-synchronous replication mode in addition to the default asynchronous mode.
Storage engine Uses the TairDB storage engine developed by Alibaba Cloud, which combines disk and memory to balance data persistence with access speed.

Use cases

ESSD/SSD-based instances are well-suited for workloads where a small fraction of the dataset is frequently accessed and the rest is accessed infrequently. Typical scenarios include:

  • Warm and cold data storage: Large datasets where only a portion of data is accessed regularly and the rest is seldom read. Disk-based storage eliminates the cost of holding rarely accessed data in memory.

  • High-capacity archives: Datasets that exceed what a standalone in-memory instance can hold, requiring hundreds of terabytes of persistent storage.

  • Redis-compatible migrations: Workloads that need Redis 6.0 API compatibility combined with disk-backed persistence, without additional integration work.

ESSD-based vs SSD-based instances

Tair offers two sub-types within this series. The right choice depends on your architecture and capacity requirements.

Item ESSD-based instance SSD-based instance
Storage medium ESSDs at performance levels PL1, PL2, and PL3. PL3 outperforms PL2 and PL1. Local SSDs
Instance architecture Standard architecture only Standard and cluster architectures
Storage capacity Custom, in increments of 10 GB Fixed capacity options
Backup and restoration Snapshot-based backup of cloud disks — faster backup and restoration speeds Physical backup; speed varies by data volume
Cost Higher than SSD-based instances of comparable specifications More cost-effective than ESSD-based instances of the same specifications

Choose ESSD-based instances if you need custom storage capacity increments, faster snapshot-based backup and restoration, or plan to scale storage independently of compute.

Choose SSD-based instances if you need cluster architecture support or want the most cost-effective option at a fixed capacity tier.

Limitations

  • Command compatibility: Not all Redis 6.0 commands are supported. Review the full list at Limits on commands supported by Tair.

  • Architecture: ESSD-based instances support only the standard architecture. Cluster architecture is available only on SSD-based instances.

  • Performance trade-off: Disk-based storage delivers approximately 60% of the throughput of Redis Open-Source Edition instances. Workloads that require sub-millisecond latency across the full dataset are better served by memory-optimized instances.

Instance specifications

For available instance types and capacity options, see ESSD/SSD-based instance specifications.

FAQ

What storage engine and Redis version do ESSD/SSD-based instances use?

ESSD/SSD-based instances run the TairDB storage engine, developed in-house by Alibaba Cloud, and are compatible with Redis 6.0. For the full list of supported commands, see Limits on commands supported by Tair.

References