Skip to content

GCP Professional Cloud Database Engineer (PCDE) - Exam Strategy

Format reminder

  • 50-60 questions, 120 minutes
  • Pass mark ~70-75%

Top traps

  1. Database product matrix (memorize):
  2. Cloud SQL: regional, MySQL/PostgreSQL/SQL Server, fully managed
  3. AlloyDB: PostgreSQL-compatible, performance-tuned, vertical+horizontal scale
  4. Spanner: global ACID, horizontal scale, SQL
  5. Bigtable: wide-column NoSQL, petabyte scale, single-key access
  6. Firestore: document, mobile/web sync
  7. Memorystore: managed Redis / Memcached

  8. Memorystore for Valkey: open-source Redis fork

  9. Migration via Database Migration Service: supports MySQL→Cloud SQL, PostgreSQL→Cloud SQL/AlloyDB, SQL Server→Cloud SQL, Oracle→Cloud SQL/AlloyDB/Spanner. Heterogeneous migrations require schema work.

  10. HA modes:

  11. Cloud SQL: regional HA (sync standby in another zone)
  12. AlloyDB: regional with read pools
  13. Spanner: regional or multi-region; multi-region is global ACID

  14. Bigtable: replication for HA + DR

  15. Cloud SQL Editions: Enterprise (default) vs Enterprise Plus (newer, better performance, longer maintenance windows, near-zero downtime maintenance).

  16. Spanner config: regional (single region) vs multi-region (us, eu, asia, nam-eur-asia1). Cost and consistency vary.

  17. Bigtable replication: app profiles, multi-cluster routing, single-cluster routing. Failover behavior depends.

  18. Cloud SQL connection options: public IP + Cloud SQL Auth Proxy, private IP via VPC, Private Service Connect endpoint. PSC is the modern recommendation.

  19. Encryption: default Google-managed; CMEK via Cloud KMS; CSEK for Cloud SQL is limited.

  20. PITR and backups: Cloud SQL PITR (1-7 days), Spanner PITR (up to 7 days), AlloyDB continuous backup. Each product has different mechanics.

  21. Consistency levels in Spanner: strong reads (default) vs stale reads (timestamp bounded, lower latency, eventual). Can request via API.

High-yield topics easy to miss

  • AlloyDB Omni (run AlloyDB on-prem)
  • Database Insights (per-database performance dashboards in Cloud SQL)
  • Spanner change streams (CDC)
  • Cloud SQL Studio (in-console SQL editor)
  • Bigtable continuous backup
  • Datastream (CDC service for replication into BigQuery / Cloud Storage / Cloud SQL)

Time management

120 / ~55 = ~2.2 min/question.

When stuck

  1. Match access pattern to product - SQL transactional → Cloud SQL/AlloyDB/Spanner; wide-column → Bigtable; document → Firestore.
  2. Match scale to product - regional small → Cloud SQL; regional large → AlloyDB; global → Spanner; PB → Bigtable.
  3. Default to managed over self-hosted on GCE.
  4. Eliminate "use [random other product]" when the access pattern doesn't fit.

Day-of logistics

120 min, ~55 questions.

After

Pass: Cert valid 2 years.

Fail: Most failures are on Database Selection (~25%) - re-review the product matrix above.

PCDE patterns

  • "Global ACID + SQL" = Spanner
  • "PostgreSQL + analytics + scale" = AlloyDB
  • "MySQL/PostgreSQL/SQL Server managed regional" = Cloud SQL
  • "Wide-column at PB scale" = Bigtable
  • "Document + mobile sync" = Firestore
  • "Managed Redis" = Memorystore
  • "MySQL→Cloud SQL migration" = Database Migration Service
  • "PITR within 7 days" = Cloud SQL PITR or Spanner PITR
  • "HA failover automatic" = Cloud SQL HA / AlloyDB regional / Spanner multi-region
  • "CDC stream" = Datastream or Spanner change streams