Filestore and Shared File Storage

Filestore provides managed network-attached file storage using the NFS protocol. Multiple machines mount the same share and see the same directory tree, with standard file semantics: partial writes, renames, locking and permissions.

That is the crucial difference from object storage. Object storage requires you to replace a whole object; a file system lets you seek to a position and modify part of a file. Software written for a file system generally cannot use object storage without being rewritten.

A large amount of existing software assumes a POSIX file system — content management systems, media processing tools, scientific applications, and older enterprise software. During a migration, a shared file system is often what makes lift-and-shift possible at all.

It also solves a specific architectural problem: several machines needing simultaneous read-write access to the same data. Persistent disks generally attach to one instance for writing; a network file share does not have that limitation.

• Service tier — determines performance and capacity range. Higher tiers deliver more throughput at higher cost.
• Capacity affects performance — throughput generally scales with provisioned capacity, so a share can be too small to be fast enough.
• Mount target — the address clients mount, reachable from within your VPC network.
• Snapshots and backups — point-in-time copies for recovery.
• Regional availability — higher tiers can offer availability across zones within a region.

• Content management systems where several application servers share an uploads directory.
• Media rendering farms where many machines read the same assets.
• Scientific and engineering applications that expect a POSIX file system.
• Home directories and shared workspaces for development environments.
• Lift-and-shift migrations of software that cannot be modified to use object storage.