RethinkDB is a persistent, industrial-strength key-value store with full support for the Memcached protocol.
RethinkDB currently is fully compatible with the Memcached protocol, allowing you to seamlessly drop RethinkDB behind any application that uses the Memcached protocol for its cache or backing data store.
More protocols will be added in future versions of RethinkDB.
While RethinkDB performs best on dedicated commodity hardware that has a multicore processor and is backed by solid-state storage, it will still deliver a performance advantage both on rotational drives and in the cloud.
Read our performance report for more details.
Get more questions answered by reading the FAQ.
RethinkDB achieves linear scalability on multicore machines by eliminating contention between cores. Distributing data across cores also allows RethinkDB to take full advantage of NUMA machines. See the performance numbers on multicore machines.
RethinkDB is up to ten times faster than competing solutions that run on solid-state drives. We test dozens of storage systems in our lab, and our transformation engine ensures the drive is always performing at peak efficiency. See the difference these innovations make to the performance on SSDs and on rotational drives for I/O-bound workloads.
We strive to measure and report performance in a clear, informative, and statistically sound manner. We measure more than a dozen workloads that span different usage patterns and hardware configurations, and publish a full performance report for every version of RethinkDB.
In most scenarios, RethinkDB operates in a sub-millisecond latency range. When higher throughput is required, the latency profile is still within reasonable bounds. See the latency histograms for more details.
RethinkDB supports tens of thousands of concurrent connections due to its sophisticated event system. See the results of the concurrency benchmark for more details.
RethinkDB operates at memory speeds on very large databases, as long as data in the most frequently accessed blocks fits in RAM. See the benchmarking results for a canonical database workload for more details.
Any application or client library written for Memcached will continue to work with RethinkDB, without modification. Deploying RethinkDB is a drop-in process; any instance running a Memcached-compatible product can be replaced by RethinkDB without any hassle.
RethinkDB supports access to raw block devices to eliminate the 1% - 10% overhead associated with file system access. RethinkDB's own consistency checks eliminate the need for a filesystem and its lengthy verification procedures following power failure.
RethinkDB's performance is excellent out of the box, but we allow a high degree of control over the database internals. We expose various configuration options like block size, garbage collection policies, parallelism control, and much more.
RethinkDB allows data to be split across multiple files or block devices. In unpredictable storage systems (e.g. some RAID controllers or Amazon EBS), this ensures that a single volume's performance constraints will not impact overall performance.
Any data stored in RethinkDB can be extracted to an open-format file. Extracted data can be restored to any Memcached-compatible instance, even if data integrity is compromised.
RethinkDB maintains performance for databases that far exceed the amount of RAM on a system.
RethinkDB uses 40% less memory per page of data than many competing solutions. Since less memory is being used for bookkeeping, more data can be cached, further improving performance.
In the case of power failure, RethinkDB immediately recovers and starts responding to queries, without lengthy verification and consistency checks. The unique data layout techniques we use guarantee fault-tolerance and data consistency.
Choose between minimal durability guarantees for absolute peak performance, or strong durability guarantees for the assurance that data will always be committed to disk. See how performance is affected by various durability modes.
RethinkDB ships with tools that allow database administrators to ensure the consistency and correctness of production instances during tasks like data recovery.
In the case of partial failure of the underlying storage system, RethinkDB provides tools to recover available data from the storage device.