Scaling Global Database Reads in Google Cloud

Table of Contents

1. Challenge

   Reproduce the read-latency failure and confirm the freshness requirement

   • Deploy a Cloud SQL PostgreSQL primary instance in Europe and configure a North American test VM.
   • Populate the inventory database and measure cross-region read latency against the under-100-ms target.
   • Establish the five-second data-freshness requirement for inventory updates.
2. Challenge

   Evaluate a cross-region read replica against a Cloud Spanner migration for that requirement

   • Compare Cloud SQL cross-region read replicas and Cloud Spanner based on read performance, consistency, and implementation effort.
   • Evaluate the suitability of both options for the existing PostgreSQL workload and five-second freshness requirement.
   • Select the Cloud SQL read replica architecture for implementation and validation.
3. Challenge

   Provision the selected configuration

   • Create a Cloud SQL PostgreSQL read replica in the North American region.
   • Configure network access and database connectivity from the North American test VM.
   • Verify that the replica is operational and the inventory records are accessible.
4. Challenge

   Test the result with a cross-region write-then-read check

   • Measure regional read latency from the North American VM to the Cloud SQL replica.
   • Update an inventory record on the European primary database.
   • Execute a write-then-read test to measure when the updated record becomes visible on the replica.
5. Challenge

   Verify the observed freshness meets the requirement

   • Compare the measured replica visibility delay against the five-second freshness requirement.
   • Determine whether the freshness test meets the requirement based on the observed result.
   • Assess the selected architecture against the measured performance and freshness requirements.