Dragonfly Redis-Compatible In-Memory Data Store

DragonflyDB Ltd. builds Dragonfly, an in-memory data store designed from scratch in 2022 by co-founders Oded Poncz and Roman Gershman to remove the scaling limits of single-threaded caches such as Redis. Its legal terms are governed by Israeli law, and it raised 21 million USD from Redpoint and Quiet Capital in 2023 when version 1.0 became generally available.

Dragonfly is API compatible with Redis, Valkey and Memcached, so applications can migrate without code changes. A multi-threaded, shared-nothing architecture uses all CPU cores of a server, while SSD data tiering moves large values to local NVMe and Dragonfly Swarm manages clusters of dozens of shards. It serves caching, real-time feature stores, message queues and leaderboards, available as a self-managed build or the managed Dragonfly Cloud. Users include Instacart, Hugging Face, ShareChat, Meesho and redBus.

• Multi-threaded engine: shared-nothing design that scales vertically across all CPU cores.
• Protocol compatibility: supports Redis, Valkey and Memcached APIs with 308 commands in version 2.0.
• SSD data tiering: moves large values to local NVMe, raising effective capacity up to 8x.
• Dragonfly Swarm: manages workloads with dozens of shards of 200-400 GB each.
• Snapshot streaming: serialization without memory spikes during backups.
• Dragonfly Cloud: managed service with free start and custom enterprise quotes.

• Vietnamese e-commerce platforms handling flash-sale traffic with a larger, cheaper cache tier.
• Fintech and e-wallet apps storing sessions, rate limits and real-time risk features.
• Social and media platforms running leaderboards and news feeds at scale.

• Up to 25x more QPS than Redis in the vendor benchmark (3.97 million versus 148,000 QPS).
• About 30% better memory efficiency; a 1 TB cache costs about 6,800 USD per month versus 17,082 USD for Redis in the vendor comparison.
• Instacart reported a 70% smaller cluster and 40-50% lower latency; ShareChat cut infrastructure cost by 40%.
• Drop-in compatibility with Redis, Valkey and Memcached clients.

High-traffic apps, AI feature stores and real-time personalisation push in-memory caches to larger sizes, while Redis licensing changes have led many teams to evaluate compatible alternatives. E-commerce, ride-hailing and social platforms in Vietnam and Southeast Asia face peak-season traffic that makes cache efficiency a direct cost lever.

DragonflyDB Ltd., terms governed by Israeli law (with DragonflyDB Inc. as a US invoicing entity). Project launched in 2022; raised 21 million USD led by Redpoint and Quiet Capital in March 2023. Over 31,000 GitHub stars and more than 25 million downloads.