Skip to main content

Storage & Databases

Exam guide§2.2, 3.2

The single most-tested storage skill: the exam hands you a workload and expects one product. Two spectrums - Storage (object, file, cache) and Databases (relational, document, wide-column, analytics) - and scenario matching is what gets tested. Classify the data shape first, THEN read the scale/consistency signals.

ObjectCloud StorageGood for:Binary orobject dataSuch as:Images, mediaserving,backupsFileFilestoreGood for:NetworkAttachedStorage (NAS)Such as:LatencysensitiveworkloadsRelationalCloud SQLGood for:WebframeworksSuch as:CMS,eCommerceSpannerGood for:RDBMS +scale, HA,HTAPSuch as:User metadata,Ad/Fin/MarTechAlloyDBGood for:Hybridtransactionaland analyticalprocessingSuch as:Machinelearning,Generative AINon-relationalFirestoreGood for:Hierarchical,mobile, webSuch as:User profiles,game stateBigtableGood for:Heavy read+ write,eventsSuch as:AdTech,financial, IoTWarehouseBigQueryGood for:EnterprisedatawarehouseSuch as:Analytics,dashboardsRedisMemorystoreGood for:Automatingcomplex Redisand MemcachedtasksSuch as:Enabling highavailability,failover,patching
Google Cloud’s storage and database services grouped by data model - object, file, relational, non-relational, analytics warehouse, and Redis cache - with what each is good for and typical uses.
StartIs your datastructured?Do you need ashared file system?FilestoreCloud StorageDoes your workloadinvolve analytics?Do you need extensiveupdates and/orlow latency?BigtableNoSQL, high throughputBigQueryWarehouse, SQL queryingIs your datarelational?Do you needHTAP?AlloyDBDo you need globalscalability?SpannerCloud SQLDo you needapplication caching?MemorystoreFirestoreYESNO
Start at "Is your data structured?" and follow the YES/NO branches to one product. Unstructured splits on whether you need a shared file system (Filestore) or object storage (Cloud Storage). Structured splits on analytics, then relational, then the scale/consistency signals that separate AlloyDB, Spanner, Cloud SQL, Memorystore, and Firestore.
DECISIONWhich database?
Relational, single region, MySQL/Postgres/SQL Server→Cloud SQL
Relational + global scale + horizontal + strong consistency→Spanner
High-performance Postgres (HTAP, 4x faster)→AlloyDB
Document / mobile / web app, real-time sync, offline→Firestore
Massive wide-column, time-series, IoT, > 1 TB, low latency→Bigtable
Analytics, data warehouse, SQL over huge datasets, BI→BigQuery
In-memory cache, sub-ms reads, session store→Memorystore
Pick this when: match data shape first, then scale
GotchaRelational ≠ automatically Cloud SQL

Watch the scale words. "Global", "horizontal scaling", "strong consistency at scale", "millions of QPS with SQL" mean Spanner, not Cloud SQL. Cloud SQL scales UP (bigger instance) and out only with read replicas - it does NOT shard writes across regions.

CompareBigtable vs BigQuery - one letter, opposite jobs
BigtableOperational NoSQL: low-latency single-row reads/writes, no SQL analytics.
BigQueryAnalytical warehouse: scans huge tables with SQL. NOT for high-throughput single-row lookups or transactional writes.

If the question says "analytics/reporting/dashboard", it's BigQuery.

DECISIONSQL vs NoSQL signal words
"transactions", "JOINs", "schema", "ACID"→Relational (Cloud SQL / Spanner)
"flexible schema", "documents", "mobile sync"→Firestore
"time-series", "IoT", "petabytes", "single-key"→Bigtable
"data warehouse", "analytics", "ad-hoc SQL", "BI"→BigQuery
"reduce DB load", "sub-millisecond", "cache"→Memorystore
Pick this when: the question's vocabulary is the answer
0%0 of 147 pages studied