What Makes ScyllaDB So Fast?

ScyllaDB is a monstrously fast and scalable distributed NoSQL database designed for high throughput and low latency at scale. But what makes ScyllaDB so fast?

First, ScyllaDB uses a shard-per-core architecture, based on a framework called Seastar, making it perfect for modern servers with dozens or even hundreds of CPU cores. ScyllaDB automatically partitions the database into shards, each running on a CPU core with associated memory and storage. These shards work as independently as possible, which is called a “shared-nothing” approach. 

On the client side, ScyllaDB’s shard-aware drivers connect directly to the exact right CPU core within the node that holds your data. This further lowers latencies and increases throughputs.

Next, ScyllaDB is written in C++ so it runs efficiently, taking full advantage of your infrastructure and direct access to the low-level mechanisms of the Linux operating system. For example, ScyllaDB can optimize its IO concurrency to the exact type of fast storage you’re using. Being written in C++ also means users don’t need to tinker with heap memory tuning or worry about pauses due to garbage collection. Because ScyllaDB can provide greater performance, you require less infrastructure, which means you get direct savings.

ScyllaDB uses its own built-in CPU and IO schedulers. It can automatically prioritize its own activities due to real-time, real-world conditions. Even under the most intense workloads ScyllaDB runs smoothly without requiring frequent administrator supervision and intervention.

ScyllaDB is also fast for queries because it has a built-in row-based data cache. Frequently-used data will already be stored in memory. There’s no need for an external caching layer, saving you on complexity and infrastructure costs.

ScyllaDB distributes data across regional and global datacenters, so users can access their data locally, minimizing connection round trip times.

There’s a lot more to ScyllaDB under the hood, so we invite you to try it for yourself. Go to ScyllaDB.com to get started today.