The new IO model and scheduler provide fine-tuned balancing of mixed read/write requests based on the disk’s capabilities. This boosts throughput, reduces latency, and minimizes the impact of admin operations. It also reduces latency for real-time requests while running admin operations like scaling out.
ScyllaDB’s new adaptive caching strategy removes the performance penalty traditionally associated with large partitions – even with reversed queries. ScyllaDB can now handle even a 100GB partition with ease.
Get strong, immediately consistent schema updates, topology changes, tables and indexes, and more using the Raft consensus protocol. This eliminates schema and data conflicts, enables rapid and safe increases in cluster capacity, and provides a leap forward in manageability. Schema updates are in 5.0.
Repair-based node operations provide significant improvements for performance and data safety. If an operation fails (for example from a network hiccup), it will resume from the same point. Plus, there’s no need to run repair before or after node operations.
Eliminate the risk of unwanted data resurrection with ScyllaDB automatically tracking the completion of repair operations and purging outdated tombstones. This improves performance by allowing compaction to remove tombstones earlier and reducing read amplification.
ScyllaDB is written on Seastar, an asynchronous, non-blocking runtime that works extremely well for building highly-reliable low-latency distributed applications. It’s a perfect complement for applications written in Rust. We offer a fully asynchronous Rust driver that uses the Tokio runtime. This will be the core code unifying our drivers, with language bindings to C++, C, /C++ and Python.
* = available now