Open-source CockroachDB alternatives

TiDB is a cloud-native, open-source distributed SQL database from PingCAP. The name's Ti stands for titanium, and it is built for high availability, strong consistency and the ability to scale out as data and traffic grow, while staying compatible with MySQL so existing applications and tools can connect with little change.

Distributed transactions use a two-phase commit protocol to keep ACID guarantees even across nodes, through network partitions or node failures. It scales horizontally by adding nodes or vertically by adding resources, without downtime, and its architecture separates compute from storage so each can be adjusted independently. Built-in Raft consensus keeps multiple replicas in sync and handles automated failover. Topics also mention HTAP workloads, analytics and native vector search.

TiDB is written in Go under the Apache-2.0 license. It can be deployed on your own infrastructure or consumed as a managed cloud service, and it suits teams that have outgrown a single MySQL server and want horizontal scale without sharding by hand.

Citus is an extension that transforms PostgreSQL into a distributed database so it can keep performing as data and traffic grow. Instead of replacing Postgres, it adds capabilities to it, and because it is an extension it works with recent PostgreSQL versions and the tools and extensions people already use.

Distributed tables are sharded across a cluster of nodes to pool CPU, memory, storage and I/O, while reference tables are replicated to every node for joins and foreign keys. A distributed query engine routes work to the right shards and runs SELECT, DML and other statements in parallel, columnar storage compresses data and speeds up scans, and queries can be sent to any node. More worker nodes can be added and shards rebalanced as data grows.

Typical uses include multi-tenant applications, real-time analytics, high-throughput transactional workloads and time-series or IoT data. Citus is written in C, licensed under AGPL-3.0 and described by its authors as fully open source, and it can start as a single node and scale out later. The project links a SIGMOD paper describing how it works.

YugabyteDB is a distributed SQL database that is compatible with PostgreSQL. It combines the relational features of traditional databases with the scale-out design of NoSQL systems, and is intended for cloud-native OLTP applications that need correct data and at least one of scalability, tolerance to failures or global distribution.

Its SQL layer, YSQL, reuses the PostgreSQL query layer and so supports most PostgreSQL features, including data types, stored procedures, triggers and extensions. Its transaction model follows the design of Google Spanner, using Raft consensus for replication and hybrid logical clocks for distributed ACID transactions, with snapshot, serializable and read committed isolation. Reads are strongly consistent by default and can be tuned to read from followers and replicas.

The database can be configured to withstand failures at the level of a disk, rack, node, zone, region or whole cloud, recovering on its own, and the topics mention multi-region, multi-cloud and Kubernetes deployments. The project is written mainly in C and C++, the repository lists its license as Other, and documentation covers architecture, roadmap and how to build applications.

OceanBase is a distributed relational database developed by Ant Group. It runs on a cluster of commodity servers and uses the Paxos protocol and a distributed architecture to provide high availability and scalability, without depending on a specific hardware architecture.

It is compatible with MySQL, which eases migration from MySQL databases, and supports hybrid transactional and analytical processing, so one engine can serve both OLTP and OLAP workloads. The project now describes itself as a unified, multi-model database for the AI era, with topics that include vector search and full-text search. Cloud-native operation is supported through a Kubernetes operator.

You can try it by deploying a stand-alone instance with an all-in-one package on Linux, starting a Docker container from images on Docker Hub, quay.io or ghcr.io, or running it on Kubernetes with ob-operator. OceanBase is written in C++ and licensed under Apache-2.0, with a developer guide for compiling the observer yourself.

YDB is a distributed SQL database for workloads that need scale without giving up consistency. It offers high availability and scalability along with strict consistency and ACID transactions, and it was designed from scratch for scalable interactive web services with an OLTP-like profile, where effective transactions across rows are essential.

The README says the project was built by people with a deep background in databases and distributed systems, with experience building a NoSQL database and a MapReduce system at a very large search company. The code base is written in C++ and described as cloud-native, with documentation, a blog and community channels on Discord and Telegram.

YDB is licensed under Apache-2.0 and can be deployed on your own infrastructure. It suits teams that want a horizontally scalable relational-style database with strong transactional guarantees, for example for large online services, and are willing to run a distributed system.