Which database does Meta use?

Meta (formerly Facebook) uses a variety of databases, optimized to handle its massive scale and performance requirements. Here are the primary databases used at Meta:

1. MySQL (with Custom Enhancements)

• Usage: MySQL serves as a core part of Meta's relational database system, particularly for managing structured data. Meta has customized MySQL significantly to meet its massive-scale needs.
• Enhancements: One major enhancement is the use of RocksDB, an embedded storage engine, to improve MySQL’s performance in handling read-heavy workloads.
• Use Case: MySQL is used for many of Meta's transactional workloads and services that require structured data.

Source:

• MySQL at Meta

2. RocksDB

• Usage: RocksDB is a high-performance key-value store that Meta developed to handle the needs of large-scale data management. It is used as a storage engine for MySQL, helping to optimize read-heavy workloads.
• Use Case: RocksDB is used to optimize access patterns and manage large volumes of data with low latency, making it a key part of Meta’s infrastructure.

Source:

• RocksDB Overview

3. Cassandra

• Usage: Meta uses Apache Cassandra, a highly scalable NoSQL database, for handling large amounts of unstructured data. It is well-suited for distributed, highly available systems, which is critical for Meta’s infrastructure.
• Use Case: Cassandra is employed in applications requiring high availability, fault tolerance, and scalability, such as Facebook Messenger and other real-time communication services.

Source:

• Cassandra at Meta

4. HBase

• Usage: Meta uses Apache HBase, another NoSQL database, for large-scale storage of structured data. HBase is designed to provide random, real-time read/write access to large datasets.
• Use Case: HBase is used for various applications that require real-time analytics and random access to massive amounts of data, such as maintaining social graph data.

Source:

• HBase Overview

5. Presto (Trino)

• Usage: Presto, an open-source distributed SQL query engine developed at Meta, is used to query large datasets stored in various data lakes and storage systems. It is built to run interactive queries on large-scale data without moving or copying it.
• Use Case: Presto is primarily used for analytics and data science queries, allowing engineers to analyze massive datasets efficiently.

Source:

• Presto Overview

Conclusion:

Meta uses a mix of MySQL with RocksDB for transactional workloads, Cassandra and HBase for distributed and unstructured data, and Presto for querying large-scale datasets. Each of these databases is optimized to handle the vast amount of data generated and processed by Meta’s platforms.

Sources:

• MySQL at Meta
• RocksDB Overview
• Cassandra at Meta
• Presto Overview