Dragonfly

Dragonfly is a Redis and Memcached compatible in-memory data store you run yourself or as the managed Dragonfly Cloud, with cloud memory billed per GB.

Dragonfly gives applications written for Redis a server that runs commands on many CPU cores in parallel. It accepts the Redis and Memcached APIs over one shared keyspace, so existing client libraries connect to it as a drop-in replacement. Its multi-threaded, shared-nothing architecture scales a single instance vertically to multi-terabyte memory. DragonflyDB Ltd., co-founded in 2022, develops the server and operates Dragonfly Cloud. The data model covers hashes, sorted sets, streams with consumer groups, and JSON documents read and updated by path. FT.CREATE builds search indexes over hashes and JSON, with text, tag, numeric, vector and geo fields. Pub/sub channels, atomic Lua scripts through EVAL and CLIENT TRACKING behave the way Redis clients expect. Snapshots write the dataset to disk, to Kubernetes persistent volumes or to S3, while SSD Data Tiering moves less-used values out of memory. Cluster mode shards the keyspace across nodes, and ACL users are limited by command and key pattern over TLS connections. Dragonfly suits teams on Redis or Amazon ElastiCache that want more throughput from each node while keeping the same API. Platform teams can run it themselves with the Dragonfly Operator on Kubernetes, which reconfigures replication when a primary fails. Teams that would rather not operate it pick Dragonfly Cloud on AWS, Google Cloud or Azure, where a replica is promoted with no Sentinel involved. The Business tier adds bring your own cloud, SSO with MFA and a custom SLA. Self-Managed places Dragonfly on a company's own cloud or on-premises hardware, with enterprise authentication and support. The common jobs are caching and session storage in front of a primary database. Ranking features keep leaderboards in sorted sets, and background workers use Dragonfly as a BullMQ queue or a pub/sub bus. Go services pair it with the Rueidis client for server-assisted client-side caching, and Django applications cache querysets through cacheops. For a team leaving ElastiCache, the same Redis commands carry over, so moving to Dragonfly is a change of server rather than a rewrite of application code.