Cloud Spanner Overview Google Professional Data Engineer GCP

  1. Home
  2. Cloud Spanner Overview Google Professional Data Engineer GCP
  • A fully managed service
  • relational database service
  • offers transactional consistency at global scale
  • automatic synchronous replication for high availability
  • Has interleaved schema – Group primary key and foreign keys together for faster access
  • Cascade on delete can be done with interleave.
  • Primary keys don’t have sequential keys and make sure its distributed
  • Secondary indexes make the query efficient


CAP Theorem

The CAP theorem is for distributed system and they can guarantee at most two qualities only:

  • Consistency: All observers see the most recent data and order of events is guaranteed
  • Availability: The system is always online and able to handle all requests
  • Partition tolerance: The system continues to operate during network disruptions


  • RDBMS can best provide CP
  • NoSQL can provide AP
  • Spanner provides all three – CAP.


Comparison with other data models

Cloud Spanner Traditional Relational Traditional Non-Relational
Schema Yes Yes No
SQL Yes Yes No
Consistency Strong Strong Eventual
Availability High Failover High
Scalability Horizontal Vertical Horizontal
Replication Automatic Configurable Configurable




  • Used For strong consistency
  • Google datacenters have atomic clocks
  • It gives Spanner nodes to determine the current time down to a very fine resolution. the margin of error in system time is about 7 ms.


Paxos consensus algorithm

  • Allows consensus to be reached in very unreliable environments
  • ideal for maintaining consensus across large bodies of data.


Read / Write

  • queries are performed using strong reads that guarantee the most recent results
  • queries may result in a slight performance hit.
  • Clients can specify an exact staleness in queries
  • the client provides a timestamp and Spanner executes the query on the most recent data relative to that timestamp.
  • Spanner APIs provide write operation
  • For write give a number of mutations to be executed,
  • each mutation affects a single cell
  • Spanner supports up to 20,000 mutations in a single transaction, and affected indexes are also counted.
  • Cloud Spanner offers two types of transaction
  • read-only transactions – non-locking operations and it ensures that the data being read is not updated from the observer’s point of view over the course of one or many reads. Any update during a read-only transaction are ignored
  • read-write transactions – locking operations so, data will be blocked until the transaction is committed. It supports rollbacks.