ScyllaDB

ScyllaDB, Inc. was launched in 2014 by Dor Laor and Avi Kivity after their work at Red Hat, opened its US headquarters in 2016 and lists offices in Sunnyvale (California), Herzliya (Israel) and Warsaw (Poland). It builds a distributed NoSQL database for data-intensive applications that need very high throughput with predictable low latency, positioned as a faster and cheaper alternative to Apache Cassandra and Amazon DynamoDB.

The database uses a close-to-the-metal, shard-per-core architecture written in C++, exposes the Cassandra CQL interface and a DynamoDB-compatible API called Alternator, and adds vector search for RAG and feature-store workloads. It is offered as the fully managed ScyllaDB Cloud and X Cloud on AWS and GCP, in the customer's own cloud account, or as self-managed ScyllaDB Enterprise on-premises or on Kubernetes. Customers shown on the site include Discord, Disney, Tencent, Palo Alto Networks, Zillow, Tripadvisor and Agoda.

• Shard-per-core engine: A C++ architecture that pins data to CPU cores for predictable low latency.
• Cassandra and DynamoDB compatibility: CQL drivers and the Alternator API ease migration from both systems.
• Vector search: Stores and searches billions of embeddings for RAG, recommendation and agentic workloads.
• Full-text search: Release 2026.3 adds a full-text index with BM25 relevance scoring.
• Tablets elasticity: Tablet-based data distribution scales clusters quickly and restores from object storage.
• Flexible deployment: Managed cloud on AWS and GCP, bring-your-own-account, or self-managed Enterprise and Kubernetes Operator.

A Vietnamese e-wallet or super app can store user profiles, sessions and transaction history for tens of millions of users with low latency. An e-commerce or content platform can run a real-time feature store and vector search for personalized recommendations. A telecom operator can replace an aging Cassandra cluster with fewer nodes and lower infrastructure cost.

Teams can run millions of operations per second with predictable single-digit millisecond latency on a smaller cluster than Cassandra-style systems, and the vendor claims around 50 percent lower cost. Existing Cassandra and DynamoDB applications can migrate with compatible APIs instead of a rewrite. Reserved 1- or 3-year plans can cut cloud costs by up to 70 percent versus on-demand, according to the pricing page, and the managed service is SOC 2 and ISO certified.

Recommendation engines, feature stores and AI agents need operational databases that serve huge volumes of reads and writes with stable latency, while cloud bills from managed NoSQL services keep rising. Digital platforms in Southeast Asia, including travel, gaming, fintech and e-commerce companies, face the same scale and cost pressure as they grow regional user bases.

ScyllaDB was launched in 2014 by Dor Laor (CEO) and Avi Kivity (CTO), opened its US headquarters in 2016 and has team members in 27 countries. Investors listed include BVP, Eight Roads, Innovation Endeavors, Qualcomm Ventures, TLV Partners, AllianceBernstein and Vertex Ventures. The core database is published on GitHub.