DragonflyDB is a modern open source in-memory database, compatible with Redis and Memcached API, no need to modify any code when migrating, and can be used as an alternative to both. Compared with traditional memory data storage, DragonflyDB provides higher throughput and cache hit ratio, lower tail-latency, and convenient vertical scalability.
DragonflyDB implements a new algorithm and data structure based on the multi-threaded and Shared-nothing architecture. Its performance is 25 times that of Redis, and a single instance supports millions of QPS.
Recently, DragonflyDB officially released version 1.0 , and the development team said it has supported the use in production environments. They have invested significant development resources in four key areas - performance, scalability, efficiency and reliability . Dragonfly 1.0 fully supports the most common data types and commands of Redis, as well as features such as snapshots, master-slave replication, and high availability.
According to the official benchmark data , the performance of Dragonfly is much higher than that of Redis.
Dragonfly uses an innovative hash table structure called dashtable to minimize memory overhead and latency. Dragonfly also uses bitpacking and denseSet to compress memory data, making its memory efficiency 30% higher than Redis on average. Dragonfly uses consistent memory during snapshots, eliminating the need for overprovisioned memory typical of Redis.
For future plans, the development team mentioned that SSD storage will be used to expand the main memory while retaining Dragonfly's low-latency features, allowing users to use hardware more efficiently and reduce usage costs.