GCP Storage Architecture & Data Systems Hub 💽
Google Cloud provides an integrated storage portfolio spanning high-throughput globally distributed object storage, ultra-low-latency block volumes, POSIX-compliant managed Network Attached Storage (NFS), and automated petabyte-scale migration pipelines.
Service Catalog & Architecture Map
| Service | Category | Protocol & Performance Profile | Reference Guide |
|---|---|---|---|
| Google Cloud Storage (GCS) | Object Storage | HTTP/REST, sub-second latency across all classes (Standard, Nearline, Coldline, Archive), Bucket Lock WORM. | Google Cloud Storage (GCS) |
| Persistent Disk & Hyperdisk | Block Storage | Attached block volumes, Zonal/Regional synchronous mirroring, Hyperdisk Extreme (up to 500k IOPS) and Hyperdisk ML. | Persistent Disk & Hyperdisk |
| Filestore | Managed File (NFS) | POSIX-compliant NFSv3 / NFSv4.1 shared storage, ReadWriteMany (RWX) multi-VM packing, and sub-millisecond file access. | Filestore Deep Dive |
| Storage Transfer Service | Migration & Sync | Serverless, high-throughput data migration between AWS S3, Azure Blob, on-prem filesystems, and Cloud Storage. | Storage Transfer Service |
| GKE Storage CSI & FUSE | Kubernetes Storage | CSI drivers dynamically provisioning Persistent Disks, Filestore shares, and streaming GCS object datasets into Pod memory. | GKE Storage CSI |
Architectural Storage Selection Framework
flowchart TD Req["Data Storage Requirement"] --> AccessPattern{"What is the access pattern and filesystem requirement?"} AccessPattern -->|"Unstructured Files / Media / Backups / AI Training"| GCS["Use Google Cloud Storage (GCS)"] AccessPattern -->|"Direct Low-Latency VM Operating System / Database Block"| Block["Use Persistent Disk or Hyperdisk"] AccessPattern -->|"POSIX Shared Filesystem / ReadWriteMany (NFS)"| NFS["Use Cloud Filestore (NFSv3/v4.1)"] AccessPattern -->|"Terabyte-to-Petabyte External Migration"| Migration["Use Storage Transfer Service"] GCS --> Streaming{"Need local POSIX mount in Kubernetes?"} Streaming -->|"Yes"| GCSFuse["Mount via Cloud Storage FUSE CSI"] Streaming -->|"No"| GCSDirect["Direct REST / gcloud API"] Block --> IOPS{"Extreme IOPS / ML throughput required?"} IOPS -->|"Yes (Up to 500k IOPS / 10 GB/s)"| Hyperdisk["Hyperdisk Extreme / Hyperdisk ML"] IOPS -->|"Standard General Purpose"| StandardPD["Persistent Disk Balanced (pd-balanced)"]
1. Object Storage vs Block Storage vs Shared File
- Google Cloud Storage (GCS): Scalable, durable (), globally accessible object storage. All classes (even Archive) deliver millisecond first-byte latency, unlike AWS Glacier which requires multi-hour restore jobs.
- Persistent Disk & Hyperdisk: Block devices attached to specific VMs or GKE nodes. Cannot be mounted across multiple VMs simultaneously (except in read-only mode).
- Filestore: Shared Network Attached Storage (NFS). Allows hundreds of Compute Engine VMs or GKE pods to concurrently mount the same directory with ReadWriteMany (RWX) semantics.
Related Knowledge Hubs
- GCP — Master Google Cloud Platform architectural catalog
- GKE Storage CSI Driver — Kubernetes storage classes and volume snapshots
- GKE Cloud Storage FUSE CSI — AI/ML dataset streaming from GCS
- AWS Storage Comparison — Comparison with AWS S3, EBS, and EFS