Dragonfly delivers high performance, scalability, and sub-millisecond latency, optimizing cloud applications with cutting-edge infrastructure and efficiency.

Dragonfly is a high-performance, scalable, and sub-millisecond data store designed as a drop-in replacement for Redis. It is built to take advantage of the power of modern cloud hardware to deliver optimal performance while cutting costs. Dragonfly aims to address the data needs of modern cloud applications by achieving sub-millisecond latency, ensuring faster and more efficient user experiences. The approach of Dragonfly is towards simplicity. This design will make an infrastructure simple with the avoidance of any component which is not needed. This way, the utilization of modern hardware to the full extent using vertical scaling can be achieved, and it gives a more efficient, reliable in-memory data store as compared to traditional ones. It allows developers to enjoy efficiency while avoiding limitations usually found in legacy systems.

Discover Dragonfly's Key Features

Performance

Dragonfly delivers amazing performance with millions of QPS per instance, all due to non-contending, multi-threaded processing that achieves the maximum possible throughput on the same hardware. Its unique caching algorithm results in higher hit rates and offers much better protection for primary databases. The platform maintains sub-1ms P99 latency consistently and thus minimizes tail latency, with reliable performance under heavy loads. Dragonfly is designed to natively integrate with Kubernetes, and it provides an easy-to-use Kubernetes operator that makes deployment and management easy. Therefore, it fits well in the modern cloud environment, providing developers with an efficient solution for high-performance, low-latency data demands. With high throughput, low latency, and streamlined management, Dragonfly is an ideal solution for modern application infrastructure.

Scalability

Dragonfly can scale up the vertical simple way, making the scaling of one node very easy, cheap, and predictable than Redis Cluster or other multi-node solutions. In addition, this multi-threaded, shared-nothing architecture supports TB-sized workloads by scaling vertically all the way to 1TB of memory per instance without a complex multi-node management system. This design removes the requirement of resharding, because Dragonfly takes care of distributing data automatically across the CPUs. Thus, the operation becomes straightforward, and it does away with the scaling complexities associated with it. This focus on vertical scaling, using modern cloud hardware, allows scaling to be as easy and efficient as possible for big applications. The result is an amount of flexibility in growing the demand for data and with less overhead and maintenance than usual when scaling using other methods.

Efficiency

Dragonfly is built with the highest standards of efficiency; it consumes memory 30 to 60% better than the conventional data stores. With its novel structures of data, it makes effective use of any available compute resource, directing unused power towards compressing for even better usage in memory. Running tasks asynchronously in the platform gives the highest use of CPU cycles for better throughput on the whole. Hardware resources used by Dragonfly help the same workloads fit into smaller instances, which minimizes hardware expenses. It also supports efficient snapshotting with predictably low memory usage during the process, thus ensuring minimal performance impact. Snapshots are completed up to 30 times faster than with Redis, making it a more efficient solution for data management. Advanced memory optimization, asynchronous task execution, and faster snapshotting position Dragonfly — as an ideal choice for organizations looking to reduce operational costs while maintaining high performance.

Simplicity

Dragonfly focuses on simplicity. It supports Redis and Memcached API compatibility, so organizations can enjoy increased speed and performance without changing their existing code. It features a single binary that can be easily deployed via Docker, Kubernetes, or any other preferred environment, streamlining setup and management. Dragonfly also supports local development and CI/CD workflows, ensuring a smooth transition for teams already using Redis or Memcached. Native Prometheus is supported, without any requirement for an additional agent, to simplify observability. Also, it supports cloud snapshotting, where the platform would allow users to write and read snapshot files from S3-compatible solutions in the first place. This combination of API compatibility, streamlined deployment options, simple monitoring, and integrated cloud snapshotting makes Dragonfly an easy-to-use solution that fits flawlessly into existing infrastructures.

Oded Poncz - Co-Founder & CEO

“Accelerating modern cloud applications by providing the most performant, scalable, sub-millisecond data store in the galaxy.”