ScyllaDB

active

by ScyllaDB Inc. · since 2015 · also known as Scylla

C++ rewrite of Cassandra with shard-per-core design; also speaks the DynamoDB API.

wide-columnkey-valuedatabase enginesource-availablemature

Overview

ScyllaDB re-implements the Cassandra data model and CQL protocol in C++ on the Seastar framework, pinning one shard to each CPU core so it avoids JVM pauses and lock contention. In practice that means far fewer, larger nodes for the same throughput and tighter p99 latencies, which is why teams such as Discord moved trillions of messages onto it. Tablets-based elastic sharding, Raft-managed schema and topology, and the Alternator DynamoDB-compatible API make it a migration target from both Cassandra and DynamoDB. Since the 2025 releases the single product is source-available rather than AGPL: free up to 10 TB and 50 vCPUs, commercial beyond that.

Strengths & trade-offs

Strengths

  • +Shard-per-core architecture delivers low, stable p99 latency without JVM GC pauses
  • +Drop-in CQL compatibility plus a DynamoDB-compatible API (Alternator)
  • +Tablets give fast, automatic rebalancing when scaling out or in
  • +Typically needs several times fewer nodes than Cassandra for the same load
  • +Workload prioritization lets OLTP and batch jobs share a cluster

Trade-offs

  • −No longer open source: production use beyond 10 TB / 50 vCPUs requires a commercial license
  • −Same query-first modelling limits as Cassandra; no joins
  • −Some Cassandra features and edge behaviours differ, so migrations need testing
  • −Smaller community and managed-service choice than Cassandra

Great for

  • +Chat and messaging history at very high volume
  • +Low-latency feature stores for ML
  • +Cassandra or DynamoDB migration targets
  • +IoT and telemetry ingestion
  • +Ad-tech and personalisation lookups

Think twice for

  • −Teams that require an OSI-approved open-source license
  • −Relational or join-heavy data
  • −Small single-node workloads

Capabilities

Native or official first-party support · 11 of 25 tracked

✓Secondary indexes
✓JSON
✓Change data capture
✓Materialized views
✓TTL / expiry
✓Durable persistence
✓Compression
✓Encryption at rest
✓Role-based access
✓Online schema changes
✓Streaming ingest
–Transactions
–Joins
–Full-text search
–Vector search
–Geospatial
–Triggers
–Stored procedures
–Pub/Sub
–Columnar storage
–In-memory
–Time travel
–Branching
–Offline sync
–Runs in WebAssembly

Scale & consistency

Comfortable data size
Massive · Multi-PB
Horizontal scaling
Native
Multi-region
Active active
Transactions
Limited ACID
Consistency
Tunable, Eventual
Schema
Strict
Architecture
Distributed, Multi primary, Client server
Query languages
CQL, DynamoDB API

Tablets rebalance data automatically as nodes are added; petabyte-scale clusters are publicly documented.

Deployment

Self-hosted Vendor-managed cloud Serverless Embedded Edge Kubernetes operator

available on · AWS · Google Cloud

Ecosystem

compatible withCassandra (CQL), DynamoDB
apis & protocolsCQL native protocol, Alternator (DynamoDB-compatible HTTP API)
client languagesJava, Python, Go, Rust, C++, C#, JavaScript, Scala
used byDiscord, Comcast, Zillow

Alternatives to ScyllaDB

Compare all→
Apache CassandraWide-Column55.9

Masterless wide-column database built for always-on writes across data centers.

scylladb vs apache-cassandra →
Amazon DynamoDBKey-Value63.3

AWS's serverless key-value and document store with single-digit-millisecond reads at any scale.

scylladb vs amazon-dynamodb →
DataStax Astra DBWide-Column11.7

Serverless Cassandra-as-a-service with vector search and a JSON Data API, now owned by IBM.

scylladb vs datastax-astra-db →
Amazon KeyspacesWide-Column2.4

Serverless, CQL-compatible Cassandra service on AWS with no clusters to manage.

scylladb vs amazon-keyspaces →
Google Cloud BigtableWide-Column33.3

Google's managed wide-column store, the original Bigtable, with HBase API and SQL access.

scylladb vs google-cloud-bigtable →
AerospikeKey-Value40.4

Flash-optimized distributed key-value and document database for predictable sub-millisecond latency at scale.

scylladb vs aerospike →

ScyllaDB FAQ

What is ScyllaDB?

ScyllaDB is a wide-column database by ScyllaDB Inc., first released in 2015. ScyllaDB re-implements the Cassandra data model and CQL protocol in C++ on the Seastar framework, pinning one shard to each CPU core so it avoids JVM pauses and lock contention. In practice that means far fewer, larger nodes for the same throughput and tighter p99 latencies, which is why teams such as Discord moved trillions of messages onto it. Tablets-based elastic sharding, Raft-managed schema and topology, and the Alternator DynamoDB-compatible API make it a migration target from both Cassandra and DynamoDB. Since the 2025 releases the single product is source-available rather than AGPL: free up to 10 TB and 50 vCPUs, commercial beyond that.

Is ScyllaDB open source?

Not in the OSI sense. ScyllaDB is source-available: you can read the code, but the license restricts some uses, typically offering it as a competing managed service. License: ScyllaDB Software License Agreement 1.1.

Does ScyllaDB support ACID transactions?

Partly. ScyllaDB offers limited ACID guarantees (for example, single-record or single-partition transactions), with tunable, eventual consistency.

Does ScyllaDB scale horizontally?

ScyllaDB scales horizontally out of the box, with built-in sharding or partitioning. It is comfortable at massive scale (Multi-PB). It supports active-active multi-region writes.

How can I run ScyllaDB?

You can self-host it and use the vendor's managed cloud. Managed offerings include ScyllaDB Cloud. It is available on AWS and Google Cloud.

What is ScyllaDB best for?

ScyllaDB is a strong fit for chat and messaging history at very high volume; low-latency feature stores for ML; and cassandra or DynamoDB migration targets. Think twice for teams that require an OSI-approved open-source license.

What are the best alternatives to ScyllaDB?

Common alternatives to ScyllaDB include Apache Cassandra, Amazon DynamoDB, DataStax Astra DB, Amazon Keyspaces and Google Cloud Bigtable. The right choice depends on your workload, scale and hosting; WhichDB's smart match ranks them against your requirements.

How popular is ScyllaDB?

ScyllaDB ranks #36 of 844 databases on the WhichDB Score (45.9/100), and #2 among wide-column databases. The score combines public adoption signals such as package downloads, GitHub activity, job posts and developer discussion.