Skip to main content

Filestore

Exam guide§2.2
Filestore

Fully managed NFS file storage (Network Attached Storage) - a shared POSIX file system you mount on Compute Engine VMs or GKE clusters. Unlike Cloud Storage (object, reached by API) Filestore gives applications a real file-system mount point, so it fits lift-and-shift apps that expect files and directories, latency-sensitive workloads, and shared scratch space for HPC or rendering.

FilestoreNFS shareCompute Engine VMCompute Engine VMGKE clusterNFSv3
One Filestore share is mounted over NFSv3 by many clients at once - a fleet of Compute Engine VMs and a GKE cluster all read and write the same file system, so a render fleet or migrated app shares one directory tree.

Pick a service tier by performance and capacity needs (Basic HDD/SSD for general file sharing, higher tiers for more throughput and scale). Performance and capacity tune independently, which is what gives file-based workloads predictable, repeatable speed.

GotchaFilestore vs Cloud Storage

If the workload needs a mounted file system / NFS / shared directory across VMs, it's Filestore. If it needs to store and serve blobs by key/URL with no file system, it's Cloud Storage.

NumbersFilestore capabilities
  • NFSv3 - supports any NFSv3-compatible client; no plug-ins or client-side software to install or maintain.
  • Scale-out performance and hundreds of terabytes of capacity per instance.
  • File locking built in, so multiple clients can safely share the same files.
  • Grow or shrink an instance's capacity or performance as needs change.
  • Mount points for both Compute Engine and GKE.

Use cases

Any workload that expects a POSIX file system shared across many machines.

Use caseWhy Filestore fits
Application migrationOn-prem enterprise apps that need a file-system interface lift-and-shift without rewrites.
Media renderingRender fleets of Compute Engine VMs all mount one share, so artists collaborate on the same files.
Electronic Design Automation (EDA)Batch jobs across thousands of cores with large memory needs; files stay universally accessible.
Data analyticsLow-latency file ops for financial/environmental models; a persistent shared layer avoids loading data to local drives.
Genomics processingBillions of data points per genome need speed, scale, and predictable pricing for the performance.
Web contentHosting providers and CMSes (e.g. WordPress) serve shared web content from one file system.
0%0 of 147 pages studied