Google Cloud PCA Study Guide
Section 3: Storage, Backup, and Data Transfer
Service selection guide focused on why an organization would choose each storage, backup, or transfer option
PCA study frame For PCA-style storage questions, the right answer usually comes from matching the data access pattern to the storage type: object, block, file, archive, backup, or transfer. The most common traps are confusing object storage with file systems, choosing the coldest storage class without considering retrieval/minimum duration, and treating HA as a backup strategy. |
Services and Concepts Covered
Object storage and policy | Block and file storage | Backup and transfer |
|---|---|---|
Cloud Storage / GCS | Persistent Disk / PD | Backup and DR Service |
Standard storage class | Standard PD | Storage Transfer Service |
Nearline | Balanced PD | Transfer Appliance |
Coldline | SSD PD | Data retention / lifecycle design |
Archive | Extreme PD | Backup and recovery design |
Lifecycle policies | Hyperdisk | |
Object versioning | Local SSD | |
Retention policies / Bucket Lock | Filestore | |
Google Cloud NetApp Volumes | ||
Google Cloud Managed Lustre |
Fast PCA Decision Matrix
Use this table when a practice question gives you a business requirement and asks what the architect should recommend.
If the organization needs... | Likely PCA answer |
|---|---|
Need generic object storage for unstructured data at massive scale | Cloud Storage |
Need hot/frequently accessed object data with no minimum duration | Standard storage class |
Need lower-cost object storage accessed about monthly | Nearline |
Need cold online storage accessed rarely and retained at least 90 days | Coldline |
Need long-term archive or compliance data retained at least one year | Archive |
Need objects to move/delete automatically as they age | Lifecycle policies |
Need recovery from object overwrite or accidental deletion | Object versioning |
Need immutable/WORM-style retention or legal hold | Retention policies / Bucket Lock |
Need durable block storage attached to VMs or GKE | Persistent Disk or Hyperdisk |
Need lowest-cost durable block storage for non-latency-sensitive workloads | Standard PD |
Need general-purpose VM disk cost/performance balance | Balanced PD |
Need higher-performance durable block storage for databases | SSD PD |
Need provisioned very high IOPS on supported VM shapes | Extreme PD |
Need modern configurable block performance or newest machine support | Hyperdisk |
Need fastest local scratch/cache disk and data can be recreated | Local SSD |
Need managed NFS shared file storage | Filestore |
Need enterprise file storage with NetApp-style features | Google Cloud NetApp Volumes |
Need parallel high-throughput file system for AI/HPC | Managed Lustre |
Need managed backup, recovery, monitoring, and restore workflows | Backup and DR Service |
Need online/cloud-to-cloud/on-prem recurring data transfer | Storage Transfer Service |
Need to move huge data faster by shipping hardware | Transfer Appliance |
Service Deep Dives
Each card is written in PCA language: what the service/concept is, why an organization would use it, the exam trigger, and the common wrong turn.
Object storage
Cloud Storage / GCS What it is: Google Cloud object storage for unstructured data such as files, backups, images, logs, data lakes, and static assets. Why an organization would use it: Use Cloud Storage when an organization needs highly durable object storage that can scale without managing file servers or block devices. It is the default answer for object data, backups, archives, static website assets, analytics landing zones, and cross-region storage patterns. PCA exam trigger: Unstructured data, massive scale, global access, backup/archive target, static content, data lake landing zone, or lifecycle-based cost optimization. Watch out: Do not pick Cloud Storage when the application requires a mounted POSIX/NFS file system or low-latency block storage attached to a VM. |
Standard storage class What it is: The Cloud Storage class for frequently accessed data with no minimum storage duration. Why an organization would use it: Use Standard when data is hot, actively served, frequently processed, or accessed unpredictably. It is the safest default when access patterns are unknown because lower-cost cold classes add minimum-duration and retrieval considerations. PCA exam trigger: Hot data, active website assets, frequent analytics input, frequently accessed application content, or uncertain access pattern. Watch out: Do not over-optimize into Nearline/Coldline/Archive when the data will be read often or deleted quickly. |
Nearline storage class What it is: Lower-cost Cloud Storage for data accessed less frequently, with a 30-day minimum storage duration. Why an organization would use it: Use Nearline for backups, monthly-access data, older logs, or data that still needs quick online access but is not hot. It balances lower storage cost with occasional retrieval. PCA exam trigger: Data accessed roughly monthly, backup data that may need online restore, or lowering cost without going fully cold. Watch out: Do not use Nearline for temporary objects or high-read workloads because retrieval and minimum-duration charges can erase savings. |
Coldline storage class What it is: Lower-cost Cloud Storage for infrequently accessed data with a 90-day minimum storage duration. Why an organization would use it: Use Coldline for disaster recovery copies, quarterly-access data, older backups, or compliance data that is rarely read but still needs online availability. PCA exam trigger: Rare access, DR copy, older backup tier, data retained for at least 90 days, or lower storage cost for cold online data. Watch out: Do not use Coldline for data that is likely to be retrieved frequently or deleted before 90 days. |
Archive storage class What it is: Lowest-cost Cloud Storage class for long-term archive data with a 365-day minimum storage duration. Why an organization would use it: Use Archive for long-term regulatory retention, legal archives, historical backups, and data kept for years with very rare access. It keeps data online through the same Cloud Storage API while optimizing storage cost. PCA exam trigger: Long-term archive, compliance retention, data stored for at least a year, rarely accessed but still online. Watch out: Do not use Archive for short-term retention, frequent restore testing, or workloads with unpredictable reads. |
Lifecycle policies What it is: Rules that automatically transition, delete, or otherwise manage Cloud Storage objects based on age, version, storage class, or metadata conditions. Why an organization would use it: Use lifecycle policies to reduce manual operations and automate cost optimization. An org can move objects from Standard to colder classes as they age, delete old temp data, or clean up noncurrent versions. PCA exam trigger: Automated cost tiering, retention cleanup, object aging, deleting old data, or reducing manual storage management. Watch out: Do not recommend a manual process or scheduled script when Cloud Storage lifecycle management satisfies the requirement natively. |
Object versioning What it is: A Cloud Storage feature that keeps older object generations when objects are overwritten or deleted. Why an organization would use it: Use object versioning when an organization needs protection from accidental overwrite/deletion or wants object-level rollback. It is useful for critical files, configuration artifacts, and datasets where recovery from bad writes matters. PCA exam trigger: Recover accidentally deleted/overwritten objects, keep previous object generations, or protect against application mistakes. Watch out: Do not enable versioning without lifecycle cleanup or costs can grow from retained noncurrent versions. |
Retention policies / Bucket Lock What it is: Controls that prevent Cloud Storage objects from being deleted or modified before a required retention period expires. Why an organization would use it: Use retention policies and Bucket Lock for compliance, legal hold, WORM-style requirements, and protecting regulated data from early deletion. This is governance, not just backup. PCA exam trigger: Compliance retention, immutable storage, legal hold, audit evidence, or data must not be deleted before a set time. Watch out: Do not confuse retention with lifecycle deletion. Retention blocks deletion; lifecycle policies can automate deletion when allowed. |
Block storage
Persistent Disk / PD What it is: Durable network-attached block storage for Compute Engine VMs and GKE workloads. Why an organization would use it: Use Persistent Disk when a VM needs durable boot or data disks and Hyperdisk is not required or not available for the machine type. It is common for general VM workloads that need data to survive VM stop, restart, or failure. PCA exam trigger: Durable VM disk, boot/data disk, block storage for Compute Engine, or GKE persistent volume where performance needs are not extreme. Watch out: Do not use Persistent Disk for shared NFS semantics. Use Filestore/NetApp Volumes for shared file access. |
Standard PD What it is: HDD-backed Persistent Disk for lower-cost sequential workloads. Why an organization would use it: Use Standard PD when cost matters more than latency/IOPS and the workload is sequential or throughput-oriented, such as some batch jobs, logs, or infrequently accessed disk data. PCA exam trigger: Lowest-cost durable block storage for non-latency-sensitive VM data. Watch out: Do not choose Standard PD for database or latency-sensitive random I/O workloads. |
Balanced PD What it is: General-purpose Persistent Disk option balancing cost and performance. Why an organization would use it: Use Balanced PD for normal VM boot disks and general application data where SSD-level cost is not justified but HDD is too slow. It is often the practical default for everyday VM workloads. PCA exam trigger: General-purpose VM disks, cost/performance balance, normal boot disks, standard app workloads. Watch out: Do not pick Balanced PD for the highest IOPS/throughput requirements if SSD PD, Extreme PD, or Hyperdisk is a better fit. |
SSD PD What it is: SSD-backed Persistent Disk for higher-performance block storage. Why an organization would use it: Use SSD PD for databases, latency-sensitive workloads, and applications needing stronger IOPS/latency than Balanced or Standard PD. PCA exam trigger: Database-like workload, higher IOPS, latency-sensitive VM disk, stronger durable block performance. Watch out: Do not choose SSD PD purely for cheapest storage when the workload is cold or sequential. |
Extreme PD What it is: High-performance Persistent Disk option designed for demanding IOPS workloads. Why an organization would use it: Use Extreme PD when the workload needs provisioned high IOPS and strong performance on supported VM shapes, especially high-end databases or performance-sensitive enterprise apps. PCA exam trigger: Very high IOPS block storage, enterprise database performance, or explicit provisioned performance requirement. Watch out: Do not overuse Extreme PD when Balanced/SSD/Hyperdisk meets requirements more cost effectively. |
Hyperdisk What it is: Google Cloud durable block storage family designed for newer Compute Engine performance and scalability requirements. Why an organization would use it: Use Hyperdisk when an organization needs advanced durable block storage with configurable performance, high throughput/IOPS options, or support for newer machine series where Persistent Disk may not be available. It is increasingly the modern block storage answer. PCA exam trigger: High-performance durable block storage, latest machine families, configurable IOPS/throughput, AI/ML data loading, or modern VM disk design. Watch out: Do not assume Persistent Disk is always available or always the highest-performance choice on modern machine series. |
Local SSD What it is: Ephemeral, physically attached SSD storage on the host machine. Why an organization would use it: Use Local SSD when performance is more important than durability and the application can rebuild or replicate data elsewhere. It is good for caches, scratch space, temp processing, and high-performance local working sets. PCA exam trigger: Highest local I/O, scratch/cache/temp data, fault-tolerant processing, or data can be regenerated. Watch out: Do not use Local SSD for data that must survive VM deletion, host failure, or maintenance events. Keep durable copies elsewhere. |
File storage
Filestore What it is: Fully managed NFS file storage for Compute Engine, GKE, VMware Engine, on-prem clients, and some serverless use cases. Why an organization would use it: Use Filestore when applications require shared file semantics, POSIX-style access, NFS mounts, lift-and-shift file shares, or GKE ReadWriteMany-style patterns. It avoids managing your own NFS server. PCA exam trigger: Shared file system, NFS, legacy app needs file share, GKE shared persistent storage, or multiple clients need the same files. Watch out: Do not choose Cloud Storage just because it stores files if the application requires mounted file system semantics. |
Google Cloud NetApp Volumes What it is: Fully managed enterprise file storage based on NetApp capabilities. Why an organization would use it: Use NetApp Volumes for enterprise file workloads that need familiar NetApp features, high-performance file shares, snapshots/replication/data management capabilities, or migration of on-prem NetApp-style workloads to Google Cloud. PCA exam trigger: Enterprise file workload, NetApp migration, advanced file data management, high-performance NFS/SMB-style needs, or preserving on-prem storage capabilities. Watch out: Do not pick NetApp Volumes for simple object storage, cheap archive, or basic VM boot disks. |
Google Cloud Managed Lustre What it is: Fully managed parallel file system optimized for AI, ML, and HPC workloads. Why an organization would use it: Use Managed Lustre when large-scale compute jobs need very high-throughput, low-latency shared access to massive datasets. It is especially relevant for GPU/TPU-heavy AI/ML training, checkpointing, and HPC workloads where storage throughput can bottleneck expensive compute. PCA exam trigger: AI/HPC workload, parallel file system, massive dataset, high-throughput shared POSIX access, keeping accelerators saturated. Watch out: Do not use Managed Lustre for normal app file shares where Filestore or NetApp Volumes is simpler and cheaper. |
Backup, recovery, and migration transfer
Backup and DR Service What it is: Managed backup and recovery service for Google Cloud workloads with centralized management, monitoring, reporting, and recovery workflows. Why an organization would use it: Use Backup and DR Service when an organization needs managed protection for VMs, disks, databases, file systems, VMware workloads, and backup operations with governance and recoverability. It is the architecture answer when backup is operationalized, not just copied somewhere. PCA exam trigger: Centralized backup management, protected recovery points, RPO/RTO planning, workload restore, business continuity, or disaster recovery controls. Watch out: Do not treat a storage copy alone as a full backup strategy. PCA expects recovery objectives, testing, retention, immutability, and operational management. |
Storage Transfer Service What it is: Managed service for moving data between object/file storage systems, including S3, Azure Blob, Cloud Storage, on-prem storage, URLs, HDFS, and file systems. Why an organization would use it: Use Storage Transfer Service when an organization needs reliable, repeatable online transfers into or within Google Cloud without building custom migration tooling. It can support migrations, syncs, data lake ingestion, and cross-cloud transfer patterns. PCA exam trigger: Online transfer, scheduled migration, S3/Azure/on-prem to Cloud Storage, file transfer to Filestore, large data movement over network. Watch out: Do not build custom scripts for large recurring transfers when Storage Transfer Service matches the source/target and reliability requirements. |
Transfer Appliance What it is: A high-capacity physical device that is shipped to move very large datasets into Google Cloud. Why an organization would use it: Use Transfer Appliance when network transfer would take too long, bandwidth is limited, or uploading would disrupt production traffic. It is a practical answer for massive one-time migrations where shipping encrypted hardware is faster. PCA exam trigger: Petabyte-scale migration, poor/limited bandwidth, network upload too slow, physical data shipping is faster or more reliable. Watch out: Do not choose Transfer Appliance for small or recurring online transfers when Storage Transfer Service is simpler. |
Design patterns and exam concepts
Data retention / lifecycle design What it is: The architecture practice of matching storage class, retention period, deletion policy, immutability, and access pattern to business requirements. Why an organization would use it: Use retention/lifecycle design to control cost while meeting compliance and operational needs. PCA questions often test whether you can reduce cost without breaking restore, audit, or deletion requirements. PCA exam trigger: Cost reduction over time, compliance retention, delete after X years, keep data immutable, or archive while preserving restore ability. Watch out: Do not optimize only for lowest storage price. Retrieval cost, minimum duration, access latency, compliance, and recovery testing matter. |
Backup and recovery design What it is: The architecture practice of defining how data is protected, restored, tested, retained, and governed. Why an organization would use it: Use backup/recovery design when the organization needs business continuity, disaster recovery, protection from deletion/corruption/ransomware, and confidence that recovery works. The design should map to RPO, RTO, regions, replication, retention, and test procedures. PCA exam trigger: RPO/RTO, DR plan, restore testing, accidental deletion, ransomware protection, regional failure, business continuity. Watch out: Do not assume high availability equals backup. HA keeps services running; backups help recover from deletion, corruption, and operational mistakes. |
Common PCA Confusion Points
Confusion point | How to think about it |
|---|---|
Cloud Storage vs Persistent Disk | Cloud Storage is object storage accessed through APIs. Persistent Disk/Hyperdisk is block storage attached to VMs or GKE nodes. |
Cloud Storage vs Filestore | Cloud Storage stores objects and is not a mounted POSIX file system. Filestore provides managed NFS for apps that expect file system semantics. |
Standard vs Nearline vs Coldline vs Archive | The colder the class, the lower the storage price generally becomes, but minimum duration and retrieval considerations matter more. |
Lifecycle policies vs retention policies | Lifecycle policies automate transitions/deletion. Retention policies prevent deletion/modification before the retention period expires. |
Object versioning vs backup | Object versioning protects object generations in a bucket, but backup design also covers RPO/RTO, restore testing, immutability, monitoring, and operational processes. |
Persistent Disk vs Hyperdisk | Both are durable block storage. Hyperdisk is the modern higher-performance/configurable family and is recommended for newer machine/performance scenarios. |
Persistent Disk vs Local SSD | Persistent Disk/Hyperdisk is durable network-attached block storage. Local SSD is ephemeral host-attached storage for scratch/cache/high-performance temporary data. |
Filestore vs NetApp Volumes | Filestore is managed NFS for common file-share needs. NetApp Volumes is stronger for enterprise file workloads needing NetApp data management features. |
Filestore vs Managed Lustre | Filestore is for general managed file shares. Managed Lustre is for AI/HPC parallel file-system throughput at large scale. |
Storage Transfer Service vs Transfer Appliance | Storage Transfer Service moves data online. Transfer Appliance moves very large data through shipped hardware when network transfer is too slow. |
High availability vs backup | HA keeps applications available during infrastructure failure. Backups recover from deletion, corruption, ransomware, and operational mistakes. |
Cram Summary
- Object storage = Cloud Storage. Block storage = Persistent Disk/Hyperdisk/Local SSD. File storage = Filestore/NetApp/Managed Lustre.
- Standard is for hot data. Nearline is 30-day minimum. Coldline is 90-day minimum. Archive is 365-day minimum.
- Lifecycle policies automate cost optimization and deletion. Retention policies enforce compliance and prevent early deletion.
- Object versioning helps recover overwritten/deleted objects but needs lifecycle cleanup to manage cost.
- Persistent Disk and Hyperdisk are durable. Local SSD is fast but ephemeral.
- Use Hyperdisk when the question emphasizes newer machine series, configurable performance, high throughput, or advanced block storage needs.
- Use Filestore when an app needs a mounted NFS file share, especially for legacy apps or GKE shared file access.
- Use Managed Lustre for AI/HPC when storage throughput must keep expensive accelerators busy.
- Use Backup and DR Service when the requirement is managed backup/recovery operations, not just putting copies in a bucket.
- Use Storage Transfer Service for online migrations/syncs. Use Transfer Appliance when physical shipping beats network upload.
- Lowest cost is not always the right answer. Retrieval, minimum duration, durability, RPO/RTO, and operational complexity are PCA traps.
Sources used: Google Cloud Professional Cloud Architect certification page; Google Cloud docs for Cloud Storage storage classes, Persistent Disk, Hyperdisk, Local SSD, Filestore, NetApp Volumes, Managed Lustre, Backup and DR Service, Storage Transfer Service, and Transfer Appliance.