Dragonfly - Modern Memcached Alternative

Performance

Millions of QPS per instance

Non-contending, multi-threaded processing mean much higher throughput on the same hardware. Read this benchmark report for the details.

Higher hit rates

Dragonfly’s unique caching algorithm means higher hit rates and better protection for your primary database.

Lower tail latency

Consistent sub 1ms P99 latency. Read this benchmark report for the details.

Simple vertical scaling

Vertically scaling a single node is quicker, cheaper, and more predictable than Redis Cluster and other multi-node approaches.

TB sized workloads

Dragonfly’s multi-threaded, shared-nothing architecture scales vertically up to 1TB of memory on each instance.

No resharding

Dragonlfy automatically shards data between CPUs, eliminating the complexity around resharding.

30-60% better memory utilization

Dragonfly’s unique data structures allow for surplus compute to be put towards compression, delivering far better memory utilization.

Asyncronous

Dragonfly runs tasks asynchronous, allowing you to squeeze more cycles out of each CPU.

Lower hardware costs

Dragonfly utilizes all the hardware resources, allowing you to run the same workloads on a smaller instances.

Efficient Snapshotting

Predictably low memory usage during snapshotting. Snapshots complete up to 30X faster when compared to Redis.

Redis/Memcached API-compatible

Already using Redis or Memcached? Start enjoying increased speed and performance without touching your code.

Single binary

A single binary that can be deployed via Docker, K8, or anywhere else you choose.

Local dev, CI/CD

Already using Redis or Memcached? Start enjoying increased speed and performance without touching your code.

Simple Monitoring

Native support for Prometheus, no agent required.