Dragonfly vs Redis: Performance Claims, Benchmark Debate, and Architectural Insights

In mid‑year, a former Google and Amazon engineer launched the open‑source memory data cache system Dragonfly , written in C/C++ and distributed under the Business Source License (BSL).

Benchmark results suggest Dragonfly may be the world’s fastest in‑memory storage system, supporting Memcached and Redis protocols while delivering higher query performance and lower runtime memory consumption.

Compared with Redis, Dragonfly reportedly achieves a 25× performance boost under typical workloads, processes millions of requests per second on a single server, and uses 30% less memory in a 5 GB storage test.

Within two months of release, Dragonfly earned 9.2 K GitHub stars and 177 forks, drawing attention alongside other Redis‑compatible systems such as KeyDB and Skytable.

In response, Redis co‑founder and CTO Yiftach Shoolman, along with Redis Labs architects, published an article titled “Does Redis Need a New Architecture After 13 Years?” presenting their own Redis 7.0 vs. Dragonfly benchmark where Redis showed 18%–40% higher throughput.

The Redis team argues that Dragonfly’s benchmark methodology—comparing a single‑core Redis instance to a multi‑threaded Dragonfly instance—does not reflect real‑world Redis deployments. They instead compare a 40‑shard Redis 7.0 cluster on AWS c4gn.16xlarge instances to Dragonfly’s largest test instance.

Redis also outlines its architectural principles: running multiple Redis processes per VM for linear scaling, limiting each process size (≤25 GB, ≤50 GB with Redis on Flash), and emphasizing horizontal scaling for elasticity, cost‑effectiveness, high throughput, NUMA friendliness, and avoiding storage throughput limits.

The article concludes that while Redis respects innovative ideas from projects like Dragonfly, it will continue to uphold its core multi‑process, no‑shared‑state design, which offers optimal performance, scalability, and resilience for real‑time in‑memory data platforms.