Announcing Redis 8.10: Compact Hash, JSONPath extensions, performance improvements, & more

Redis 8.10 targets memory efficiency and throughput with compact hashes, a new internal encoding that shares field names across hashes with identical schemas. This change is most beneficial for workloads like user profiles, sessions, or feature stores where many hashes repeat the same fields. The memory savings, up to 50%, come without requiring application changes, though operators can opt into automatic template conversion for eligible hashes. The trade-off is that compact hashes work best when field sets are stable; dynamic or frequently changing schemas may not see the same gains.

The introduction of `HIMPORT` for bulk hash ingestion further improves throughput by reducing network and server-side processing overhead. Clients send field names once, then stream only values, enabling up to 2× higher loading throughput. This is particularly useful for ETL workloads or bulk imports where large volumes of similarly structured data are ingested. However, `HIMPORT` requires client-side preparation, so existing applications may need updates to take full advantage of the performance boost.

JSONPath extensions in Redis 8.10 add arithmetic, filtering, and aggregation capabilities, allowing applications to offload more data processing to Redis. This reduces the need to transfer large JSON documents to the client for post-processing, lowering network overhead and simplifying application logic. The new syntax includes functions for strings, arrays, and aggregations, but complex queries may still require careful tuning to avoid performance pitfalls, especially with deeply nested documents.

Stream optimizations in Redis 8.10 include `MAXCOUNT` and `MAXSIZE` options for `XREAD` and `XREADGROUP`, which limit the total number of messages or reply size across multiple streams. This prevents unexpectedly large responses when reading from many streams or streams with large entries. While useful for controlling memory usage during reads, these limits may require adjustments to application logic if existing code assumes unbounded responses.

Incremental backup and restore in Redis 8.10 addresses a long-standing operational challenge for large deployments. By allowing coordinated cluster backups with lower impact on running workloads, it improves reliability without requiring downtime. However, the feature may not eliminate the need for traditional backups in all cases, particularly for deployments with strict recovery point objectives or compliance requirements.