Filestore instances are fully managed file servers on Google Cloud that can be connected to a number of client types:
- Compute Engine VMs
- Google Kubernetes Engine (GKE) clusters
- External datastores such as Google Cloud VMware Engine
- On-premises machines
- Cloud Run services
Once provisioned, you can scale the capacity of your instances according to need without any downtime.
Why Filestore?
Google Cloud offers three main types of data storage: block, file, and object storage.
As a type of persistent file storage, Filestore supports multiple concurrent application instances accessing the same file system simultaneously.
For Google Kubernetes Engine users, for example, Filestore provides multiple reader, multiple writer access, letting you mount your GKE PersistentVolumes as read-write by many nodes.
Filestore offers a versatile alternative to block storage products, such as Persistent Disk, which support only limited options for multiple writer access to a disk.When compared to object storage, such as Cloud Storage FUSE, while that product does offer some file system semantics, it lacks some of the more robust characteristics of file storage provided by Filestore. Some examples of capabilities supported by Filestore and not Cloud Storage FUSE include the following:
- POSIX compliance
- Hard links and file locking
- Concurrency control for multiple writes on the same object
For more information, see the following resources:
- Compare the relative advantages of block, file, and object storage.
- Review the storage options for HPC workloads in Google Cloud.
- Learn more about Filestore support for GKE.
- Learn more about using Custom performance with GKE to scale IOPS independently of capacity.
- Learn more about the limitations of Cloud Storage FUSE.
- Learn about storage options and access modes for GKE clusters.
Service tiers
Filestore offers multiple service tiers that vary in capacity, performance, and features. Each service tier is tailored for specific use cases:
- Zonal tier: Optimized for HPC, batch compute, media rendering, and localized workloads requiring high throughput and low latency.
- Regional tier: Designed for mission-critical workloads requiring continuous availability and regional resilience.
- Multishares for GKE (Enterprise): Optimized for GKE workloads that require high availability and multishares.
- Basic tiers (legacy): Suitable for basic file sharing and software development.
For details, see Service tiers.
Protocol support
Filestore supports the following file system protocols:
| Protocol | Supported service tiers | Highlights |
|---|---|---|
| NFSv3 | All service tiers |
|
| NFSv4.1 | zonal, regional, and enterprise service tiers |
|
For help understanding which protocol may be right for you, see About supported protocols.
Connectivity
Filestore instances can connect to any clients that are on the same VPC network, including Shared VPC networks. You can also connect to clients on remote networks, such as an on-premises machine, using Cloud VPN or Cloud Interconnect.
Networking
For information related to Filestore networking requirements, see the following resources:
- Filestore networking and IP resource requirements
- Configure NFS ports on client VMs
- Configure firewall rules
- Create an instance on a Shared VPC network in service projects
- Supported file system protocols
Performance
Custom performance lets you define the performance settings for your Filestore instances independently of the specified capacity. This means you can optimize your Filestore instances for your workload requirements and scale your file shares to meet the demands of your applications without resizing the capacity.
Custom performance is available for regional and zonal service tiers. In addition, GKE workloads can dynamically adjust performance for zonal and regional tiers without downtime.For details on performance settings, limits, and testing see performance.
Data protection
The following sections discuss Filestore instance data protection.
Encryption at rest
By default, Filestore automatically encrypts your data at rest. The durable storage behind each Filestore instance is encrypted with system-defined keys that are managed by Google.
When you delete a Filestore instance, Google discards the encryption information used by the instance, rendering the data irretrievable as per the description in