Partitioning and High-level Architecture
This lesson gives a brief overview of BigTable's architecture and its data partitioning scheme.
A single instance of a BigTable implementation is known as a cluster. Each cluster can store a number of tables where each table is split into multiple Tablets, each around 100–200 MB in size.
- A Tablet holds a contiguous range of rows.
- The table is broken into Tablets at row boundaries.
- Initially, each table consists of only one Tablet. As the table grows, multiple Tablets are created. By default, a table is split at around 100 to 200 MB.