Domain 2 of 6 · Chapter 2 of 5

Storage Systems

Unlock the complete study guide + 1,040 practice questions across 16 full exams.

Bundled into the existing Professional Cloud Architect premium course — no separate purchase.

Included in this chapter:

  • Read the access shape before the price tag
  • Cloud Storage: classes, lifecycle, and retention
  • Block and file storage for VMs and clusters
  • Choosing the right managed database
  • Exam-pattern recognition

Choosing the storage family by access shape

DimensionCloud Storage (object)Persistent Disk / Hyperdisk (block)Filestore (file)Managed databases
Access unitWhole object by nameRaw block devicePOSIX files over NFSQueryable records
Who attachesAny client over API/HTTPSOne VM at a time (regional PD spans 2 zones)Many VMs / GKE pods at onceApp clients via driver/API
Mutable in placeNo, rewrite whole objectYesYesYes (rows/documents)
Durability scopeEleven-nines, regional or multi-regionZonal, or regional (2-zone sync replica)Zonal or regional tierService-managed replication
Typical useBackups, media, data lake, archiveVM boot + data disks, databases on VMsLift-and-shift NFS apps, HPC, GKE sharesTransactional / analytic data
Cost leverStorage class + lifecycle tieringDisk type + provisioned IOPS/throughputService tier + capacityInstance size / on-demand vs provisioned

Decision tree

What is the access shape?object / block / file / recordsBlobs by nameCloud Storageclass by read cadenceOne VM diskSurvive VM stop?durable vs ephemeralShared filesFilestoremany VMs / GKE, NFSQueryable recordsDatabase (below)NoLocal SSDscratch only, fastestYesPersistent Disk / Hyperdiskregional PD for zone HAHyperdisk tunes IOPSAnalytic or operational?scan large data vs serve rowsAnalyticBigQueryOperationalRelational? Global + strong consistency?NoSQL key/docBigtable (wide-column) / Firestore (doc)RelationalCloud SQL (regional) / Spanner (global,99.999%)
Cheat sheet

Cheat sheet

  • Pick the storage family by access shape before anything else
  • Cloud Storage classes change cost, not speed or durability
  • Match the Cloud Storage class to read cadence and its minimum duration
  • Bucket location type is separate from storage class
  • Automate tiering and deletion with Object Lifecycle Management
  • Lifecycle SetStorageClass only moves objects colder
  • Lifecycle actions are asynchronous and lag the trigger
  • Lock a retention policy with Bucket Lock for WORM immutability
  • Object holds freeze a single object regardless of bucket policy
  • Cloud Storage objects are immutable; rewrite to change them
  • Local SSD is ephemeral; never store data of record on it
  • Use a regional Persistent Disk to survive a zone failure
  • Hyperdisk provisions performance independently of capacity
  • Use Filestore when many clients must share one POSIX file tree
  • Pick the Filestore tier by capacity and zonal vs regional availability
  • Cloud SQL is the default managed relational database for a regional app
  • Choose Spanner for globally-scaled, strongly-consistent relational SQL
  • Use Bigtable for high-throughput single-keyed NoSQL workloads
  • Use Firestore for document data with real-time sync and offline support
  • Use BigQuery for analytics, not for transactional row updates
  • Self-managed database files belong on durable block storage
  • Cloud SQL high availability means a REGIONAL instance with synchronous cross-zone replication

Unlock with Premium — includes all practice exams and the complete study guide.

Also tested in

References

  1. Filestore overview
  2. Cloud Storage introduction
  3. Compute Engine: about disks
  4. Hyperdisk
  5. Cloud SQL introduction
  6. Spanner documentation
  7. Bigtable overview
  8. Firestore documentation
  9. BigQuery introduction
  10. Cloud Storage classes
  11. Object Lifecycle Management
  12. Bucket Lock and retention policies
  13. Local SSD
  14. High availability with regional Persistent Disk
  15. Cloud Spanner Service Level Agreement