Cloud Bigtable

Cloud Bigtable is Google Cloud's fully managed, wide-column NoSQL database service. It is the same technology that powers some of Google's core services — Search, Maps, and Gmail. Bigtable is designed for massive analytical and operational workloads that require consistent sub-10ms latency at petabyte scale.

---

What Is Cloud Bigtable?

Cloud Bigtable is a high-performance, low-latency NoSQL database designed for:

• Massive scale — handles petabytes of data with billions of rows.
• Consistent low latency — single-digit millisecond reads and writes.
• High throughput — millions of operations per second.
• Seamless scaling — add or remove nodes without downtime.
• HBase compatible — supports the Apache HBase API, making migration from on-premises HBase straightforward.

---

Data Model

Bigtable uses a wide-column data model:

Concept	Description
Row key	The unique identifier for a row. Data is sorted lexicographically by row key.
Column family	A group of related columns. Must be defined at table creation time.
Column qualifier	The specific column within a family. Can be created dynamically.
Cell	The intersection of a row key, column family, and column qualifier. Contains a value and timestamp.
Timestamp	Each cell can store multiple versions of a value, identified by timestamp.

Example: IoT Sensor Data

Row key: sensor-001#2024-01-15T10:30:00
Column family: readings
  Column qualifier: temperature -> 22.5
  Column qualifier: humidity -> 45.2
Column family: metadata
  Column qualifier: location -> "London"
  Column qualifier: type -> "environmental"

---

Row Key Design

The row key is the most important design decision in Bigtable. It determines:

• Data distribution across nodes.
• Read efficiency — Bigtable can efficiently scan contiguous rows.
• Hotspot avoidance — poor key design causes all writes to hit the same node.

Best Practices

1. Avoid monotonically increasing keys — timestamp-only keys create write hotspots.
2. Use composite keys — combine entity ID with timestamp: sensor-001#2024-01-15T10:30:00.
3. Hash or reverse keys — reverse domain names or hash user IDs to distribute writes.
4. Design for your read patterns — related rows should be adjacent for efficient range scans.

---

Clusters and Replication