Professional Cloud ArchitectManage implementationMedium

A gaming company is launching a new multiplayer online game. The game backend runs on Google Kubernetes Engine (GKE) and interacts with a global database. During peak hours, the database experiences high contention, leading to slow response times and a poor user experience. The development team wants to offload read traffic from the primary database to improve performance and scalability without significantly re-architecting the application. Which Google Cloud service should you recommend to implement a read replica strategy effectively?

  1. ABigtable for high-throughput, low-latency NoSQL operations.
  2. BMemorystore for Redis for caching frequently accessed data.
  3. CCloud Spanner for its global, horizontally scalable nature.
  4. DCloud SQL read replicas for offloading read operations.
Show answer & explanation

Correct answer: D. Cloud SQL read replicas for offloading read operations.

Cloud SQL read replicas are specifically designed to offload read traffic from a primary database instance, improving performance and scalability for read-heavy workloads without requiring significant application re-architecture, directly addressing the scenario's requirements.

Why the other options are wrong

  • A. Bigtable is a NoSQL database for analytical and operational workloads, not a direct read replica solution for a transactional global database (implied by 'global database' and 'high contention').
  • B. Memorystore for Redis is a caching service, good for reducing database load, but it's not a read replica and requires application changes to implement caching logic.
  • C. Cloud Spanner is a globally distributed database, but migrating to it would be a significant re-architecture, not just offloading read traffic from an existing database.

Cloud SQL Read Replicas

Cloud SQL read replicas are copies of a primary Cloud SQL instance that handle read queries, distributing the read workload and improving the performance and availability of the primary instance.

  • Supports MySQL, PostgreSQL, and SQL Server.
  • Asynchronous replication from primary instance.
  • Ideal for read-heavy applications to scale performance.

Memory trick: When the main book is too busy, make copies for everyone to read.

More Manage implementation questions