● Beta — KubeEra is in active development. Profiles are live; AI-perception measurement is rolling out.
KubeEra
S

ScyllaDB

Data
Open source · NOASSERTION

ScyllaDB is a NoSQL database written in C++ on top of the Seastar framework — a shard-per-core, shared-nothing async runtime built to squeeze maximum throughput out of modern multi-core hardware.

updated 2026-07-01

What it is

ScyllaDB is a NoSQL database written in C++ on top of the Seastar framework — a shard-per-core, shared-nothing async runtime built to squeeze maximum throughput out of modern multi-core hardware. It's wire-compatible with Apache Cassandra (and also offers a DynamoDB-compatible API), so existing CQL drivers, tooling, and data models generally work without rewrites. The pitch is simple: same data model as Cassandra, drastically lower latency and higher throughput per node because it avoids the JVM and manages memory/threading itself instead of leaning on the OS scheduler and garbage collector.

Who builds it and why

Built by ScyllaDB Inc. (formerly Cloudius Systems), a commercial company, not a foundation-governed project. CNCF maturity is not publicly available because it isn't a CNCF project — this is a vendor-led open-core database. The company sells a managed cloud offering and enterprise features on top of the open-source core, which is the standard "open source funnel to commercial support/hosting" model. 241 contributors is healthy for a systems-level C++ database but reflects a company-dominated contributor base rather than a broad community-governed one — worth knowing when you're evaluating how decisions get made.

Production readiness signal

15,630 GitHub stars and continuous commits (last commit recorded 2026-07-01) indicate an actively maintained codebase, not an abandoned experiment. It has real production adoption at companies with serious throughput needs (Discord, Comcast, Zillow, among public case studies). That said:

  • License field returns NOASSERTION on GitHub — meaning the repo's license isn't cleanly machine-detected. ScyllaDB has historically used AGPL for its core with some components under other terms. Don't assume permissive licensing; check the actual LICENSE file per component before you build a commercial product on it.
  • Latest release version is not publicly available in this dataset — pin and verify the exact release you're deploying, don't assume "latest" from docs matches what you're pulling.
  • Founded date not publicly available here, but the project and company have been around long enough (public since ~2015) to have gone through multiple major version cycles and real operational hardening.

Who should use this

Teams already running Cassandra who are CPU- or latency-bound and want a drop-in replacement without re-architecting the data model. Also a good fit if you need predictable P99 latency at high write throughput (time-series, IoT telemetry, ad-tech event stores, messaging backends) and have the ops maturity to run a distributed database yourself or via ScyllaDB Cloud. If your team already understands Cassandra's data modeling constraints (partition keys, no joins, eventual consistency tradeoffs), the migration curve is short.

Who should NOT use this

  • You need strong relational guarantees, joins, or ad-hoc queries — this is still a wide-column store; the query limitations of Cassandra apply here too.
  • You're small-scale and single-node is enough — the whole value prop is multi-core, multi-node efficiency; running it on a laptop-sized workload wastes its design.
  • Your team has zero Cassandra/CQL experience and no ops bandwidth — the operational model (compaction strategies, tombstones, repair, consistency levels) has a real learning curve regardless of which implementation you pick.
  • You need a CNCF-governed, vendor-neutral project for procurement or compliance reasons — this is a single-vendor commercial product with open-source components, not a foundation project.

Alternatives

  • Apache Cassandra — the JVM-based original; larger community, slower per-node performance, more battle-tested at extreme scale across many independent operators.
  • DataStax Astra/Enterprise — Cassandra-based, adds proprietary tooling and support, more enterprise-polished but pricier and more vendor lock-in.
  • YugabyteDB / CockroachDB — if you actually need SQL and distributed ACID transactions instead of a wide-column model, these are the closer fit despite different query semantics.

Pricing

Core ScyllaDB is open source (self-managed, free to run). ScyllaDB Inc. monetizes via ScyllaDB Cloud (managed hosting, usage-based pricing) and ScyllaDB Enterprise (additional features + support, commercial license). Exact enterprise/cloud pricing is not publicly available — it's quote-based.

Frequently asked

What is ScyllaDB?+
ScyllaDB is a NoSQL database written in C++ on top of the Seastar framework — a shard-per-core, shared-nothing async runtime built to squeeze maximum throughput out of modern multi-core hardware.
Who builds ScyllaDB?+
Built by ScyllaDB Inc. (formerly Cloudius Systems), a commercial company, not a foundation-governed project. CNCF maturity is not publicly available because it isn't a CNCF project — this is a vendor-led open-core database.
Is ScyllaDB production ready?+
15,630 GitHub stars and continuous commits (last commit recorded 2026-07-01) indicate an actively maintained codebase, not an abandoned experiment. It has real production adoption at companies with serious throughput needs (Discord, Comcast, Zillow, among public case studies).
Who should use ScyllaDB?+
Teams already running Cassandra who are CPU- or latency-bound and want a drop-in replacement without re-architecting the data model.
Who should not use ScyllaDB?+
You need strong relational guarantees, joins, or ad-hoc queries — this is still a wide-column store; the query limitations of Cassandra apply here too. You're small-scale and single-node is enough — the whole value prop is multi-core, multi-node efficiency; running it on a laptop-sized workload wastes its design.
What are the alternatives to ScyllaDB?+
Apache Cassandra — the JVM-based original; larger community, slower per-node performance, more battle-tested at extreme scale across many independent operators. DataStax Astra/Enterprise — Cassandra-based, adds proprietary tooling and support, more enterprise-polished but pricier and more vendor lock-in.
How much does ScyllaDB cost?+
Core ScyllaDB is open source (self-managed, free to run). ScyllaDB Inc. monetizes via ScyllaDB Cloud (managed hosting, usage-based pricing) and ScyllaDB Enterprise (additional features + support, commercial license). Exact enterprise/cloud pricing is not publicly available — it's quote-based.