ScyllaDB
Generated by GPT-5-miniExpansion Funnel Raw 80 → Dedup 0 → NER 0 → Enqueued 0
| ScyllaDB | |
|---|---|
| Name | ScyllaDB |
| Developer | ScyllaDB Ltd. |
| Initial release | 2015 |
| Written in | C++ |
| Latest release | 2025 |
| Repository | GitHub |
| License | AGPL / commercial |
ScyllaDB is a high-performance distributed NoSQL database designed as a drop-in replacement for Apache Cassandra with a focus on low latency and high throughput. Developed by engineers with backgrounds from Facebook, Netflix, and Intel, it targets workloads similar to those supported by Amazon DynamoDB, Google Bigtable, and Apache HBase. ScyllaDB implements a shard-per-core architecture and aims to exploit modern x86 and ARM processor features and RDMA-capable network stacks.
Overview
ScyllaDB originated from efforts to reimplement ideas from Apache Cassandra and the Dynamo design paper, drawing on research from projects like Memcached, Cassandra CQL, and implementations influenced by Seastar and Linux kernel optimizations. The company providing commercial support, ScyllaDB Ltd., positions its product against services such as Amazon Web Services offerings and managed platforms like Google Cloud Platform and Microsoft Azure. ScyllaDB's roadmap has intersected with contributions from academics and engineers affiliated with University of California, Berkeley, Massachusetts Institute of Technology, and industry groups including Linux Foundation projects.
Architecture
ScyllaDB's core uses a shard-per-core design inspired by Seastar to avoid locks and context switches typical in POSIX threading. It replaces the Java Virtual Machine model used by Apache Cassandra with a C++ implementation leveraging asio-style reactors and direct memory management techniques common in Intel DPDK and RDMA implementations. Storage layout and SSTable formats bear conceptual relation to Log-Structured Merge-tree innovations and storage engines used in HBase and LevelDB, while compatibility layers support Cassandra Query Language clients and drivers used in ecosystems involving DataStax and Apache Spark. ScyllaDB integrates with orchestration systems such as Kubernetes and monitoring stacks like Prometheus and Grafana for operational telemetry.
Performance and Scalability
ScyllaDB emphasizes linear scalability across commodity servers from vendors including Dell Technologies, Hewlett Packard Enterprise, and Lenovo. Benchmarks published by the vendor and independent evaluations compare ScyllaDB to Apache Cassandra, Amazon DynamoDB, and CockroachDB, often referencing workloads modeled after trace data from Facebook, Twitter, and LinkedIn. The architecture aims to reduce tail latency seen in distributed systems studied in literature from Google and Microsoft Research, using techniques such as CPU pinning, NUMA-aware allocation motivated by Intel and AMD platform guides, and asynchronous I/O reminiscent of Netty and libuv strategies. ScyllaDB supports multi-datacenter replication strategies comparable to patterns described in the Dynamo and Spanner papers, enabling deployments across facilities in North America, Europe, and Asia.
Use Cases and Deployment
Operators deploy ScyllaDB for workloads including real-time advertising stacks used by companies like The Trade Desk, time-series ingestion similar to InfluxDB use cases, and user-profile stores at scale paralleled by Uber and Airbnb patterns. Integrations exist for streaming platforms such as Apache Kafka, ETL tools like Apache NiFi, and analytics frameworks including Apache Spark and Presto. Managed services compete with offerings from Amazon Web Services, Google Cloud Platform, and specialized vendors; ScyllaDB is offered as both self-hosted clusters on infrastructure from DigitalOcean and managed cloud services provided by the vendor. High-availability topologies mirror industry practices used by Netflix and PayPal for fault-tolerant microservice architectures.
Development and Community
Development occurs on repositories hosted on GitHub with contributors ranging from engineers formerly at Intel and Cisco to academic collaborators from Carnegie Mellon University and ETH Zurich. The project maintains issue trackers, continuous integration pipelines using services akin to Jenkins and Travis CI, and engages users through events such as KubeCon, Cassandra Summit, and meetups organized in cities like San Francisco, London, and Bangalore. Commercial support, training, and certifications are provided by ScyllaDB Ltd., while community documentation and client drivers are maintained for languages including Java (programming language), Python (programming language), Go (programming language), and C++ ecosystems.
Security and Reliability
ScyllaDB implements authentication and authorization mechanisms compatible with enterprise directories such as LDAP and Active Directory and supports encryption in transit using TLS protocols and encryption at rest via disk-level mechanisms used by vendors like NetApp and Pure Storage. Operational reliability is enhanced through repair and compaction strategies akin to those in Apache Cassandra and testing practices referencing chaos engineering principles popularized by Netflix's Chaos Monkey. Backup and disaster recovery approaches align with patterns used by Amazon S3-backed archives and snapshot tools in Kubernetes operators.