Prepare and test your skills

Prepare and test your skills

Free sample question with answers and detailed explanation below. The Professional Cloud Database Engineer practice bank has an additional 769 practice questions.Try 10 free realistic questions: instant score, no credit card, and no account required.
An e-commerce company is experiencing severe CPU bottlenecks on its single-instance Cloud SQL database during peak sales events. The workload includes high-concurrency transactional writes for order processing alongside heavy analytical read queries.
To resolve this, the engineering team needs a horizontal scaling architecture that meets these requirements:
Which database architecture and scaling strategy should the company implement?
Cloud SQL read replicas are asynchronous copies of a primary Cloud SQL instance designed to offload read-only traffic—such as reporting, analytics, and high-volume lookup queries—from the primary database server. Each read replica maintains a synchronized copy of the primary instance's data using engine-level replication.
INSERT, UPDATE, DELETE). All write operations must still go to the single primary instance, which remains a CPU and throughput bottleneck during peak sales events.The bottleneck on the Cloud SQL primary instance was caused exclusively by read-heavy analytical or catalog queries, and the single primary instance had sufficient CPU and memory headroom to handle all transactional writes.
A common misconception is treating relational read replicas like distributed multi-writer nodes. In Cloud SQL, read replicas offload read and ETL workloads from the primary cache, but horizontal write scaling requires either sharding or migrating to Cloud Spanner.
Cloud Spanner is a fully managed, globally distributed relational database designed to combine SQL semantics and ACID transactions with near-unlimited horizontal scalability. Instead of relying on a single server or manual database sharding, Spanner organizes table rows lexicographically by primary key and automatically partitions data into contiguous ranges called splits. These splits are dynamically assigned to compute nodes (or processing units), while the underlying data files reside on Colossus, Google's distributed file system.
Unlike Cloud SQL read replicas (which only scale reads) or manual Cloud SQL sharding (which adds immense operational complexity), Cloud Spanner handles horizontal write scaling natively. Compared to Bigtable, Spanner preserves full relational schemas and multi-row ACID transactions essential for e-commerce order processing.
Database sharding in Cloud SQL involves splitting a large monolithic relational database into multiple smaller, independent Cloud SQL instances called shards. Each shard stores a disjoint subset of the data based on a chosen partitioning key, such as customer ID, region, or order range, allowing both read and write operations to be spread across multiple database servers.
The organization was strictly committed to staying on a traditional MySQL, PostgreSQL, or SQL Server engine in Cloud SQL, had already hit maximum vertical CPU and memory limits, and had engineering resources dedicated to maintaining custom sharding middleware.
While Cloud SQL best practices recommend splitting oversized monolithic instances or sharding when vertical scaling is exhausted, Cloud Spanner automates this entire sharding lifecycle using splits while keeping a single unified SQL interface.
Cloud Bigtable is a petabyte-scale, wide-column NoSQL database built for high-throughput, low-latency workloads. Bigtable shards tables into contiguous row ranges called tablets stored on Colossus, separate from the cluster nodes that serve requests. Enabling autoscaling adds or removes nodes based on CPU and storage targets, while adding a second cluster enables replication and lets applications route analytical or batch reads to a separate cluster.
The workload consisted of non-relational time-series, IoT, telemetry, or personalization data requiring massive throughput and single-row operations rather than relational SQL queries and multi-row ACID guarantees.
Bigtable and Spanner share similar underlying storage concepts on Colossus (tablets vs. splits), which is why students often confuse them. However, Spanner provides the relational model, SQL engine, and distributed ACID transactions that Bigtable intentionally omits.
Answer them yourself and get your score the moment you finish. No account needed.
Full access includes
Keep the momentum going with these hand-picked practice scenarios
Last updated Validated