Google Cloud · Software · YoctoIT tech page

Spanner

The strongly consistent global database: relational SQL that scales like NoSQL — the technology that runs Google.

FOCUS · THE IMPOSSIBLE AS STANDARDACID transactions at planetary scale: the theorem sidestepped with atomic clocks
Request the Free Assessment Learn more
YoctoIT material for clients and partners · Google Cloud, BigQuery, Gemini and the other products mentioned are trademarks of Google LLC.
01 · What it is

Spanner, made clear.

Spanner is the database that wasn't supposed to exist: relational, ACID transactions, strong consistency — AND global distribution with unlimited horizontal scalability and 99.999% availability. Google uses it for Gmail and YouTube; TrueTime (atomic clocks in the data centers) is the trick that makes it possible.

99,999%
multi-region availability: five nines by SLA
Global ACID
consistent transactions across continents: no eventual compromises
∞
horizontal scale: you add nodes, you don't split the data
Spanner
OFFICIAL GOOGLE CLOUD BRANDING · SPANNER
CONSOLE REALE · CREAZIONE DATABASE SPANNER · FONTE: GOOGLE CLOUD DOCS
REAL CONSOLE · SPANNER DATABASE CREATION · SOURCE: GOOGLE CLOUD DOCS
02 · How to use it well

The things that make the difference.

How it works

Global, mission-critical appsorders, balances, inventories
SQL & ACID
Automatic sharding
Multi-region replication
relational · elastic · planetary
TrueTimethe certain time that orders the transactions
Google's infrastructurethe same as Gmail's
The relational that scales like the web

Il caso d'uso giusto

Spanner costs like Spanner: it's needed where the scale+consistency pair is non-negotiable — otherwise AlloyDB.

Distributed schema design

Keys and interleaving designed for sharding: the design decides the performance.

Cost granularity

You start from 0.1 nodes: the entry is softer than people think.

The Postgres interface

The PG dialect reduces the cognitive lock-in: developers feel at home.

03 · In depth

Global consistency: how it really works

Spanner offers ACID transactions with external consistency at planetary scale: TrueTime (atomic clocks+GPS) timestamps transactions with bounded uncertainty, data is divided into automatically rebalanced splits, multi-region configurations write in one continent and read everywhere with a 99.999% SLA. It speaks PostgreSQL (interface) and GoogleSQL, scales by processing units (starting from 100 PUs), and change streams feed the events.

  • TrueTime — time with guaranteed uncertainty: external consistency without global locks
  • Split automatici — data rebalanced under load: the hotspot handled by the engine
  • Multi-region — zero RPO and transparent failover across regions: a 99.999% SLA
  • Interfaccia PostgreSQL — PG drivers and skills reused on the global engine
  • Processing unit — starting from 100 PUs: Spanner even for starting small
  • Change stream — every write as an event: reactive integrations and caches
04 · Numbers and lifecycle

The numbers that matter.

99,999%
the multi-region SLA: five real nines
RPO 0
across regions: data loss is not contemplated
100 PU
the entry point: no need to start huge anymore
∞
horizontal scale: you add PUs, not replicas to manage
Spanner is chosen for requirements, not for fashion: data model, keys and costs assessed by us — when the five nines are truly needed.
05 · Use cases

Where it really pays off.

Inventories and bookings

The seat sold only once, worldwide.

Payments and ledgers

Consistent balances without nightly reconciliations.

Gaming and global platforms

Millions of users, a single state of the world.

When the data can neither stop nor be wrong: Spanner exists — and we know when it's truly needed.