Google BigTable Architecture
You'll learn what BigTable is and how Google assembled foundational distributed system components to build a scalable wide-column store.
What Is BigTable?
BigTable is Google's managed wide-column store designed to handle massive datasets across thousands of machines. It's not a single piece of software—it's a distributed system built on top of other proven Google infrastructure components. Think of BigTable as a house built on a strong foundation: it relies on Google File System (GFS, now Colossus), Chubby for coordination, and SSTable files for storage.
The Building Blocks
GFS/Colossus (Storage Layer)
BigTable doesn't manage raw disks. Instead, it stores data in immutable SSTable files on GFS (Google File System) or its successor Colossus. These files contain sorted key-value pairs and live in Google's distributed file system, which handles replication and fault tolerance automatically.
Chubby (Coordination Service)
BigTable uses Chubby, a distributed lock service similar to ZooKeeper, for critical coordination tasks: discovering tablet servers, storing schema information, and managing master election. Chubby ensures only one master controls the cluster at any time.
SSTable Format
Data is stored in SSTables (Sorted String Tables)—immutable, sorted files containing key-value pairs. This format enables efficient range scans and leverages sequential disk writes for performance, similar to what you've seen in HBase and Cassandra.
How They Work Together
The BigTable master coordinates metadata operations using Chubby, while tablet servers handle read/write requests by managing tablets (ranges of rows). Writes go to memory first, then flush to SSTables stored in GFS. Reads query memory and SSTables, using Bloom filters to skip unnecessary file reads—just like HBase.
Key Takeaway: BigTable demonstrates how a powerful wide-column store emerges from combining distributed file storage (GFS), coordination services (Chubby), and efficient on-disk formats (SSTables)—a pattern that influenced HBase and other open-source systems.