Course contentsShow
← System Design
Lesson 435 of 1,91912. NoSQL: Wide-Column StoresFree lesson

HBase Regions and Region Servers

Tables are divided into regions (row ranges) that are distributed across RegionServers for horizontal scalability.

HBase Regions and Region Servers

You'll learn how HBase achieves horizontal scalability by splitting tables into regions and distributing them across RegionServers.

The Problem: Scaling Tables Horizontally

A single HBase table can store billions of rows. Storing all that data on one machine is impossible. HBase solves this by dividing tables into manageable chunks called regions and spreading them across multiple servers.

What Are Regions?

A region is a contiguous range of rows from a table, sorted by row key. Think of a phone book split into volumes: A–G, H–N, O–Z. Each volume is a region containing a subset of the sorted entries.

When a table is created, it starts with one region. As data grows, HBase automatically splits regions into smaller ones (typically when they reach 1–10 GB). Each region maintains:

  • A start row key (inclusive)
  • An end row key (exclusive)
  • All columns for rows in that range

What Are RegionServers?

A RegionServer is a worker node that hosts multiple regions. It handles:

  • All read and write requests for its assigned regions
  • Region splitting when regions grow too large
  • Communication with HDFS for actual data storage

When a client wants to read or write data, it looks up which RegionServer owns the region containing that row key, then sends the request directly to that server.

The Distribution Advantage

This design enables horizontal scaling: adding more RegionServers increases both storage capacity and throughput. Each server operates independently, serving its regions without coordinating every operation with others. If a RegionServer fails, its regions can be reassigned to healthy servers.

Key Takeaway: HBase divides tables into regions (sorted row ranges) and distributes them across RegionServers, enabling horizontal scaling where each server independently handles a subset of the table's data.