Cloud Spanner

Cloud Spanner is Google Cloud's globally distributed, strongly consistent, relational database service. It combines the structure and query capabilities of a traditional relational database with the horizontal scalability of a NoSQL system. Spanner is designed for mission-critical applications that require global distribution, strong consistency, and five-nines availability.

---

What Is Cloud Spanner?

Cloud Spanner is a fully managed, horizontally scalable relational database that provides:

• Global distribution — data is automatically replicated across regions and continents.
• Strong consistency — ACID transactions with external consistency (the strongest form of consistency, stronger than serialisable isolation).
• SQL interface — standard SQL with joins, indexes, and a familiar query experience.
• Horizontal scaling — add nodes to increase throughput and storage without downtime.
• 99.999% SLA — five-nines availability for multi-region configurations.

---

How Cloud Spanner Works

TrueTime

Spanner's secret weapon is TrueTime, a globally synchronised clock system based on GPS receivers and atomic clocks in every Google data centre. TrueTime gives Spanner a precise, bounded notion of time across the globe, enabling it to order transactions consistently without the performance penalty of traditional distributed locking.

Splits

Spanner automatically divides data into splits (chunks) based on the primary key. Each split is replicated across zones (and regions in multi-region configurations). As data grows, Spanner automatically reshards — adding or merging splits to balance load.

Replication

• Regional instances — data replicated across 3+ zones in a single region.
• Multi-region instances — data replicated across multiple regions (e.g. nam6 spans US regions, eur6 spans European regions).
• Every write is committed to a quorum of replicas before being acknowledged.

---

Schema Design

Primary Keys

Spanner uses the primary key for data distribution. Choosing the right key structure is critical:

• Avoid monotonically increasing keys (e.g. auto-increment integers). These cause hotspots because new writes always go to the same split.
• Use UUIDs, hashed keys, or composite keys to distribute writes evenly across splits.

Interleaved Tables

Spanner supports interleaved tables — a parent-child relationship where child rows are physically stored next to their parent row. This is powerful for read patterns that always access a parent and its children together:

CREATE TABLE Orders (
  OrderId STRING(36) NOT NULL,
  CustomerId STRING(36) NOT NULL,
  OrderDate TIMESTAMP NOT NULL,
) PRIMARY KEY (OrderId);