This chapter covers GCP storage solutions: Cloud Storage, Persistent Disk, Filestore, and Local SSD. These topics are core to the ACE exam, appearing in roughly 15-20% of questions across multiple domains, especially under 'Planning and Configuring Storage'. You will learn the characteristics, use cases, performance limits, and pricing models of each service, as well as how to choose between them based on application requirements. Mastery of storage options is critical because incorrect choices lead to performance bottlenecks or unnecessary costs – common traps on the exam.
Jump to a section
Imagine a large distribution company managing a network of warehouses. Cloud Storage is like a public warehouse with unlimited shelf space that anyone can access via a loading dock (HTTP API) – you can store any item (object) in boxes (buckets) and retrieve them using a unique barcode (object key). It's perfect for archival, backups, and serving static assets. Filestore is like a dedicated storage room within the warehouse, connected by a high-speed conveyor belt (NFS protocol) to a specific factory floor – it provides shared file storage for multiple machines and is ideal for legacy applications that require a file system. Persistent Disk is like a private locker assigned to a specific worker – it's a block storage device attached to a single VM, offering high performance for databases and operating systems, but it cannot be shared across multiple VMs simultaneously (except in read-only mode). Local SSD is like a tool belt worn by the worker – it's extremely fast but temporary; when the worker leaves (VM stops), the tool belt is lost. Cloud Storage Fuse allows you to mount the public warehouse as a local folder, but it's slower than a dedicated connection. Understanding these analogies helps you choose the right storage for performance, durability, and cost requirements on the ACE exam.
What Are GCP Storage Solutions?
GCP offers four primary storage services: Cloud Storage (object storage), Persistent Disk (block storage), Filestore (file storage), and Local SSD (temporary block storage). Each serves different use cases based on access patterns, performance, durability, and cost. The ACE exam tests your ability to select the right service for a given scenario.
Cloud Storage
Cloud Storage is a fully managed, scalable object storage service for unstructured data. Data is stored as objects in buckets. Buckets are globally unique names, and objects are identified by a key. Objects can be up to 5 TB in size. Cloud Storage offers four storage classes: Standard (high availability, low latency), Nearline (30-day minimum storage duration, lower cost for infrequent access), Coldline (90-day minimum), and Archive (365-day minimum, lowest cost). The exam tests these minimums and use cases: e.g., data accessed once a quarter should use Coldline.
Key features: - Object versioning: Keep multiple versions of an object; can be enabled per bucket. - Object lifecycle management: Automatically transition objects to colder classes or delete them based on age. - Consistency: Strong consistency for all operations (read-after-write, read-after-metadata-update). - Encryption: Data is encrypted at rest (AES-256) and in transit (TLS). Customer-managed encryption keys (CMEK) and customer-supplied encryption keys (CSEK) are supported. - Access control: IAM roles at bucket level, ACLs at object level (legacy), and signed URLs for temporary access. - Performance: No bandwidth throttling, but write throughput can be optimized by sharding objects across prefixes.
Persistent Disk
Persistent Disk (PD) is block storage for Compute Engine and GKE. It is durable (replicated within the zone) and can be attached to multiple VMs in read-only mode. Types: - Standard PD: Backed by HDD; cost-effective for sequential reads/writes. - Balanced PD: Backed by SSD; good balance of performance and cost. - SSD PD: Backed by SSD; highest IOPS for databases. - Extreme PD: Highest performance, provisioned IOPS up to 500,000.
Key characteristics: - Size: 10 GB to 64 TB for standard/balanced/SSD; 100 GB to 64 TB for extreme. - Max IOPS: Standard (0.75 per GB, up to 3000), Balanced (4 per GB, up to 80,000), SSD (30 per GB, up to 100,000), Extreme (provisioned up to 500,000). - Max throughput: 240 MB/s for standard, up to 1200 MB/s for SSD, 2000 MB/s for extreme. - Snapshots: Create point-in-time snapshots stored in Cloud Storage (incremental, not full copies). Snapshots can be used to create new disks or migrate across zones. - Resize: Can increase disk size without detaching, but cannot shrink. - Encryption: Similar to Cloud Storage; default Google-managed, or use CMEK/CSEK.
Filestore
Filestore is a managed file storage service for Compute Engine and GKE, using NFS v3 protocol. It provides a shared file system that can be mounted by multiple VMs. Tiers: - Basic HDD: Up to 1 TB, 100 MB/s throughput. - Basic SSD: Up to 2.5 TB, 240 MB/s throughput. - High Scale SSD: Up to 100 TB, 1.2 GB/s throughput.
Key points for exam: Filestore is for legacy applications requiring a POSIX-compliant file system; it's not serverless; it's region-based; you can create snapshots for backup.
Local SSD
Local SSD provides ephemeral block storage physically attached to the host machine. It offers very high IOPS (up to 1 million IOPS for 24 disks) and low latency, but data is lost when the VM stops or is deleted. It is ideal for cache, scratch space, or temporary data. On the exam, remember: Local SSD cannot be detached and reattached; it is not durable; it is included in the instance cost (no separate charge).
Choosing the Right Storage
Object storage: Use Cloud Storage for unstructured data, backups, archives, and static web content.
Block storage: Use Persistent Disk for OS disks, database storage, and any application that requires a local filesystem.
File storage: Use Filestore when multiple VMs need to share a filesystem via NFS.
Temporary high-performance: Use Local SSD for caches, temp tables, or high-throughput scratch space.
The exam often presents scenarios like: "You need to store 10 TB of log files accessed once a month for compliance, with minimum cost." Correct answer: Cloud Storage Nearline (30-day minimum, but logs are kept longer, so Nearline is cheaper than Standard). Wrong answer: Coldline (90-day minimum may incur early deletion costs if logs are deleted before 90 days).
Configuration Commands
Create a bucket: gsutil mb gs://my-bucket
Set storage class: gsutil defstorageclass set NEARLINE gs://my-bucket
Create a Persistent Disk: gcloud compute disks create my-disk --size=100GB --type=pd-ssd --zone=us-central1-a
Attach disk: gcloud compute instances attach-disk my-instance --disk=my-disk --zone=us-central1-a
Create Filestore: gcloud filestore instances create my-filestore --zone=us-central1-a --tier=BASIC_SSD --file-share=name=vol1,capacity=1TB --network=default
Mount Filestore: mount -o nfs <IP>:/vol1 /mnt/filestore
Local SSD is added at instance creation: gcloud compute instances create my-instance --local-ssd=interface=nvme
Interactions
Cloud Storage can be accessed from Compute Engine using gsutil or client libraries; no network egress charges for same-region access.
Persistent Disk snapshots are stored in Cloud Storage; you can create a new disk from a snapshot in any zone.
Filestore volumes can be backed up to Cloud Storage using snapshots.
Local SSD pairs well with Persistent Disk for database logs on separate disks.
Create a Cloud Storage bucket
Use `gsutil mb gs://my-bucket` or the Cloud Console. The bucket name must be globally unique and follow DNS naming conventions (3-63 characters, lowercase letters, numbers, hyphens, no underscores, cannot start or end with a hyphen). You can specify the storage class at creation (default Standard) and location type (regional, dual-region, multi-region). Regional buckets are cheaper; multi-region provides geo-redundancy. The exam tests that multi-region is for serving content globally with automatic failover.
Upload an object and set lifecycle
Upload with `gsutil cp file.txt gs://my-bucket`. To set lifecycle, create a JSON file with rules (e.g., transition to Nearline after 30 days, delete after 365 days) and apply: `gsutil lifecycle set life.json gs://my-bucket`. Lifecycle rules are evaluated once per day. The exam may ask about minimum storage duration: deleting an object before 30 days in Nearline incurs a charge equal to 30 days minus actual storage time.
Create and attach a Persistent Disk
Create a disk: `gcloud compute disks create my-disk --size=100GB --type=pd-ssd --zone=us-central1-a`. Then attach to a running instance: `gcloud compute instances attach-disk my-instance --disk=my-disk --zone=us-central1-a`. Inside the VM, format and mount: `sudo mkfs.ext4 /dev/sdb`, `sudo mount /dev/sdb /mnt/data`. Persistent Disk supports live resizing: `gcloud compute disks resize my-disk --size=200GB` while attached, but the filesystem must be expanded manually.
Create a snapshot and restore
Snapshot: `gcloud compute disks snapshot my-disk --snapshot-names=my-snapshot --zone=us-central1-a`. Snapshots are incremental and stored in Cloud Storage (no direct access). To create a new disk from snapshot: `gcloud compute disks create new-disk --source-snapshot=my-snapshot --zone=us-central1-b`. This enables disk migration across zones. Deleting a snapshot only deletes data unique to that snapshot; previous snapshots remain intact.
Mount a Filestore share
After creating a Filestore instance, get the mount point from the console or `gcloud filestore instances describe`. Then on a Compute Engine VM: `sudo mount -o nfs <IP>:/vol1 /mnt/filestore`. Ensure the VM has the `nfs-common` package installed. Filestore uses NFS v3; it does not support Windows. The exam tests that Filestore is zone-based (instances must be in the same zone or region with VPC peering).
Enterprise Scenario 1: Media Content Delivery
A streaming company stores video files in Cloud Storage (multi-region) for global access. They use Standard class for hot content, lifecycle rules to transition to Nearline after 30 days, and Archive after one year. They serve content via Cloud CDN to reduce latency. Misconfiguration: setting bucket as regional instead of multi-region causes high latency for users far from the region. They also use signed URLs for temporary access to premium content. Performance tip: use parallel composite uploads for large files (gsutil cp -o GSUtil:parallel_composite_upload_threshold=150M).
Enterprise Scenario 2: Database Storage for E-commerce
An e-commerce platform runs MySQL on Compute Engine with Persistent Disk SSD for data and logs. They use snapshots for daily backups and replication to another zone. They attached multiple disks: one for OS (standard PD), one for database (SSD PD with 10,000 IOPS), and one for logs (balanced PD). They configured disk auto-resize alerts. Common issue: running out of IOPS because they chose standard PD for database; they had to migrate to SSD. They also use Local SSD for temp tables to improve query performance, accepting the risk of data loss on instance restart.
Enterprise Scenario 3: Shared File Storage for Analytics
A research lab uses Filestore to share datasets among multiple Compute Engine VMs running Spark. They chose High Scale SSD tier for high throughput. They mount the same Filestore share on 50 VMs. Problem: Filestore has a limit of 1000 file locks per instance; concurrent writes caused contention. They optimized by using smaller files and increasing the number of instances. They also use Cloud Storage for long-term archival of results, with a lifecycle rule to delete after 5 years.
The ACE exam tests storage under Objective 2.2 'Planning and Configuring Storage'. Key areas:
- Storage class minimums: Standard (none), Nearline (30 days), Coldline (90 days), Archive (365 days). Trap: selecting Coldline for data accessed monthly – Coldline minimum is 90 days, but early deletion penalty applies. Correct: Nearline.
- Persistent Disk types and limits: Standard (0.75 IOPS/GB, max 3000), Balanced (4 IOPS/GB, max 80,000), SSD (30 IOPS/GB, max 100,000), Extreme (provisioned up to 500,000). Common wrong answer: using SSD when Balanced is sufficient, increasing cost.
- Local SSD durability: Data lost on stop/delete. Trap: thinking Local SSD persists across instance stops. Reality: only persistent disks survive stop/start.
- Filestore vs Cloud Storage Fuse: Filestore for POSIX-compliant file sharing; Cloud Storage Fuse for mounting object storage as a file system (slower, not suitable for databases).
- Snapshot behavior: Snapshots are incremental and can be used to create disks in any zone. Trap: thinking snapshots are full copies or that they are stored in the same zone.
- Encryption options: Default Google-managed, CMEK (Cloud KMS), CSEK (you provide key). CSEK keys are not stored by Google; if lost, data is unrecoverable.
- Cost optimization: Use lifecycle policies; choose regional vs multi-region based on access patterns; use Standard for hot data, Nearline for warm, Coldline for cool, Archive for cold.
- Edge case: Cloud Storage buckets can be used to host static websites. The bucket must have the same name as the domain (or use a custom domain). Use gsutil web set to set main page and error page.
- Exam trap: A question asks for 'lowest latency' storage – answer is Local SSD, not Persistent Disk SSD. But Local SSD is ephemeral.
- Elimination strategy: If the scenario mentions 'shared file system', eliminate Cloud Storage and Persistent Disk; answer is Filestore. If 'block storage for database', eliminate Cloud Storage and Filestore; answer is Persistent Disk. If 'unstructured data', eliminate others; answer is Cloud Storage.
Cloud Storage storage classes: Standard (no min), Nearline (30 days), Coldline (90 days), Archive (365 days).
Persistent Disk IOPS limits: Standard 0.75/GB (max 3000), Balanced 4/GB (max 80k), SSD 30/GB (max 100k), Extreme provisioned up to 500k.
Local SSD is ephemeral – data lost on VM stop/delete; not for durable data.
Filestore uses NFS v3; only accessible from within same VPC (or peered) in same region.
Snapshots are incremental; stored in Cloud Storage (not directly accessible).
Use lifecycle policies to transition objects to colder classes or delete automatically.
Persistent Disk can be resized live (increase only); filesystem must be extended manually.
Cloud Storage supports versioning, object holds, and retention policies.
For shared file systems, use Filestore; for block storage, use Persistent Disk; for object storage, use Cloud Storage.
Encryption: default Google-managed, CMEK (Cloud KMS), CSEK (customer-supplied keys – lost keys = lost data).
These come up on the exam all the time. Here's how to tell them apart.
Cloud Storage
Object storage; no filesystem; accessed via HTTP/HTTPS
Unlimited capacity (per bucket, objects up to 5 TB)
Durable and replicated across zones/regions
Supports lifecycle policies for cost optimization
Best for backups, archives, static content
Persistent Disk
Block storage; attached as a raw disk to a VM
Size up to 64 TB per disk
Durable within a zone; snapshots enable cross-zone recovery
High IOPS options for databases
Best for operating systems, databases, and applications needing a filesystem
Filestore
Managed NFS file server
POSIX-compliant, supports file locking
High throughput (up to 1.2 GB/s)
Accessible only within VPC
Best for legacy applications requiring shared file systems
Cloud Storage Fuse
Mounts Cloud Storage bucket as a filesystem using FUSE
Not fully POSIX-compliant (no file locking)
Lower performance (limited by object storage latency)
Accessible from any VM with internet (or VPC)
Best for simple file access to bucket contents
Mistake
Cloud Storage buckets are available only in a single region.
Correct
Buckets can be regional, dual-region, or multi-region. Multi-region buckets automatically replicate data across multiple regions within a continent (e.g., US, EU).
Mistake
Persistent Disk snapshots are full copies of the disk.
Correct
Snapshots are incremental. The first snapshot is a full copy; subsequent snapshots only store changes. Deleting an older snapshot does not affect newer ones; data is consolidated.
Mistake
Local SSD persists when the VM is stopped.
Correct
Local SSD data is ephemeral. It is lost when the VM is stopped or deleted. Only persistent disks survive stop/start events.
Mistake
Filestore can be accessed over the internet.
Correct
Filestore uses NFS v3 and is accessible only from within the same VPC network (or peered networks) in the same region. It is not exposed to the internet.
Mistake
All Persistent Disk types support live resizing without detaching.
Correct
All PD types support live resizing (increase size) without detaching, but you cannot decrease size. After resizing, you must extend the filesystem using OS tools.
Reveal each answer, then mark whether you got it right. Score 60%+ to unlock the next chapter.
Nearline has a 30-day minimum storage duration and is designed for data accessed less than once a month. Coldline has a 90-day minimum and is for data accessed less than once a quarter. Coldline is cheaper per GB but has a higher minimum and retrieval cost. On the exam, if data is accessed monthly, choose Nearline; if quarterly, choose Coldline.
Create a snapshot of the disk, then create a new disk in the target zone using that snapshot. Finally, attach the new disk to a VM in that zone. Snapshots are globally available.
5 TB. If you need to store larger files, you must split them into components or use a different solution.
Yes, Cloud Storage provides strong consistency for all operations. After a write, the object is immediately available for reads. This is important for applications that require up-to-date data.
Local SSD data is lost when the VM is stopped. It persists only during the lifetime of the instance. If you need durable storage, use Persistent Disk.
No, Filestore supports NFS v3, which is not natively supported by Windows. Use a third-party NFS client or choose a different storage solution.
You've just covered GCP Storage Solutions — now see how well it sticks with free ACE practice questions. Full explanations included, no account needed.
Done with this chapter?