OceanBase — Distributed SQL Database for Enterprise Workloads

Introduction

OceanBase is a distributed relational database developed by Ant Group. It handles both transactional and analytical queries in a single engine, targeting financial services and large-scale enterprise environments that demand high availability and strong consistency.

What OceanBase Does

• Provides MySQL and partial Oracle SQL compatibility for migration ease
• Scales horizontally across commodity servers with automatic data sharding
• Delivers multi-tenant resource isolation for SaaS and cloud platforms
• Supports hybrid transactional and analytical processing (HTAP) in one engine
• Offers transparent failover with Paxos-based consensus and zero data loss

Architecture Overview

OceanBase partitions tables across multiple nodes using a range or hash strategy. Each partition is replicated via a Paxos group to ensure majority-commit durability. The SQL engine compiles queries into distributed execution plans, pushing predicates and aggregations to storage nodes. A compaction scheduler merges incremental writes from a MemTable (LSM-based) into baseline SSTable data, keeping read latency predictable.

Self-Hosting & Configuration

• Deploy via Docker for single-node evaluation or OBD (OceanBase Deployer) for clusters
• Minimum recommended: 4 CPU cores, 10 GB RAM, 50 GB disk per node for testing
• Configure tenant resources with CREATE RESOURCE POOL to control CPU and memory quotas
• Set primary_zone to control leader placement across availability zones
• Enable log archiving and backup with ALTER SYSTEM for disaster recovery

Key Features

• Paxos-based replication with automatic leader election and zero RPO failover
• Native multi-tenancy isolates CPU, memory, and I/O between workloads
• Online DDL allows schema changes without locking production traffic
• Built-in data compression reduces storage costs by 50-70% compared to MySQL
• TPC-C and TPC-H benchmark record holder demonstrating both OLTP and OLAP performance

Comparison with Similar Tools

• TiDB — Also MySQL-compatible distributed SQL; uses Raft instead of Paxos and separates storage (TiKV) from compute
• CockroachDB — PostgreSQL-compatible distributed SQL; stronger in global multi-region deployments but lacks Oracle compatibility mode
• YugabyteDB — PostgreSQL-compatible with Raft consensus; lighter operational footprint but smaller ecosystem
• Vitess — MySQL sharding middleware rather than a full database engine; requires more manual shard management
• StarRocks — Focused on OLAP analytics only; not designed for transactional workloads

FAQ

Q: Can OceanBase replace MySQL without code changes? A: OceanBase supports most MySQL 5.7/8.0 syntax and protocols. Applications using standard MySQL connectors typically work with minimal changes, though some MySQL-specific features like certain storage engine hints may differ.

Q: What is the minimum cluster size? A: A single-node Docker deployment works for development. Production clusters typically use three or more nodes for Paxos majority consensus and fault tolerance.

Q: How does multi-tenancy work? A: OceanBase creates isolated tenants within one cluster, each with its own resource pool (CPU, memory, disk). Tenants share the same physical infrastructure but cannot access each other's data.

Q: Is OceanBase fully open source? A: OceanBase Community Edition is released under MulanPubL-2.0, a license approved by OSI. The Enterprise Edition adds commercial features like Oracle compatibility mode and advanced tooling.

Sources

• https://github.com/oceanbase/oceanbase
• https://en.oceanbase.com/docs/oceanbase-database