Dragonfly

Outgrown Memorystore?

Same API. One node where you used to need a cluster.

Dragonfly Cloud runs on Google Cloud and speaks the API your apps already use. Its multi-threaded engine puts every core to work, so you scale up on fewer, smaller nodes instead of adding shard after shard.

Throughput in QPS
125k
3900k
SET
130k
3800k
GET
115k
4300k
SETEX
Redis
Dragonfly
QPS benchmark on AWS c6gn.16xlarge. Snapshot benchmark on AWS c6gn.4xlarge. Source.
Dragonfly triangle background

Powering data at

  • Spotify
  • AutoNation
  • SoFi
  • Expedia
  • Twilio

Where Memorystore runs out of room

Memorystore runs a single-threaded engine, so growth means more shards, more nodes, and a bigger bill. Dragonfly uses a multi-threaded, shared-nothing architecture that scales up on one node first, then out when you need it.

DragonflyDB logo

Dragonfly

  • Redis API Compatibility
  • Memcached API on the Same Instance
  • Threading Model
    Multi-Threaded & Shared Nothing
  • Memory Efficiency
    40% Less Memory with Dashtable
  • Snapshotting
    Fork-Free, No Memory Spikes
  • Scaling
    Scale Up on One Node, Then Out
  • Billed Through Google Cloud

Memorystore

  • Redis API Compatibility
  • Memcached API on the Same Instance
  • Threading Model
    Single-Threaded Core
  • Memory Efficiency
    Traditional Hash Table
  • Snapshotting
    Fork-Based
  • Scaling
    Scale Out by Adding Shards
  • Billed Through Google Cloud

Why teams move off managed caches

Scaling icon

Fewer nodes

Instacart cut its cluster size 70% and average latency 50%.

Lighting Squircle

Lower cost

Meesho cut infrastructure costs 60% after moving to Dragonfly.

Play Squircle

No code changes

Drop-in compatible with the clients you use today.

Dragonfly logo

See what you would save

Our engineers analyze your current setup and send projected savings, instance sizing, and migration effort within 5 business days.

Get a free cost auditView on Google Cloud Marketplace