The Ultra-Fast Key-Value Store
SplinterDB is an open source embedded high-performance general-purpose key-value store built from the ground up to maximize the utilization of the latest generation of hardware
No Compromise on Reads or Writes
SplinterDB delivers the update performance of update-optimized databases and the query performance of query-optimized databases by using our new data structure, the routing filter.
Designed for Speed
Based on a B-epsilon tree, SplinterDB represents a data structure breakthrough that is theoretically optimal in its performance.
Efficient Resource Utilization
Reduces CPU-intensive compactions, lowers write amplification, and maximizes bandwidth utilization of modern storage devices (e.g., NVMe Optane drives).
Use Cases
Best Performer of Its Class
Our study shows that SplinterDB outperforms a state-of-the-art key-value store by 10 times on writes and 4 times on reads using the YCSB benchmark.
SplinterDB on the Cloud
Do you have an application that needs to run SplinterDB on the Cloud? Are you going to be using NVMe drives? It's easy with Splinter DB.
SplinterDB on Optane-based SSDs
Is your key-value store positioned to take advantage of your newest device bandwidth? Our experiments show that SplinterDB can use over 90% Optane-based SSDs.