Get Your CertGoogle Cloud Professional Cloud ArchitectPlayCompete
‹ All Google Cloud Professional Cloud Architect questions

GCP Professional Cloud Architect Practice Question 4: Shared POSIX Filesystem Selection

Google Cloud Professional Cloud Architect · Test 1 · question 4 of 10

Solution Architecture Designeasy

An engineering simulation team is moving an existing HPC workflow to Compute Engine. Up to 200 worker VMs run a commercial solver that expects a POSIX file system mounted at a fixed path; every worker reads the same mesh files and writes result fragments into shared directories concurrently. The vendor will not support source changes, and the team wants a managed service rather than servers they operate themselves.

Which storage option should the architect recommend?

Select one answer
Show answer and explanation

Answer:

  • D. A Filestore instance sized for the working set, mounted over NFS by every worker at the path the solver expects

Filestore is Google Cloud's managed NFS file storage. It presents a true shared POSIX file system that many clients mount read-write at the same time, which is exactly the access pattern a solver expecting a mounted directory requires. Because Google operates the file servers, the team gets the shared namespace without running any infrastructure, and service tiers let them trade capacity against throughput for the working set.

Option A looks convenient but breaks the semantics the application depends on. Cloud Storage is an object store, and Cloud Storage FUSE emulates a file system over it: it has no support for concurrent partial writes to the same object, metadata operations such as directory listing and rename are expensive, and file locking is not the POSIX behaviour a solver assumes.

Option B misreads what multi-writer block storage does. Sharing a raw block device across instances does not give you a shared file system; an ordinary file system such as ext4 will corrupt if two instances mount it read-write, so multi-writer requires a cluster-aware file system, and the attachment limits are nowhere near 200 VMs.

Option C would technically deliver NFS, but it makes the team the file-server operator: they own patching, capacity growth, and, critically, a single VM that is an unmanaged single point of failure. The stem explicitly asks for a managed service.

Reference: Filestore overview

Practice Test 1 ›

Study guides for Google Cloud Professional Cloud Architect