File Storage Product Description

Selectel File Storage is a fault-tolerant, scalable file system for data storage. It can be connected to:

to cloud servers;
dedicated servers (except for the Chipcore Line);
Managed Kubernetes cluster nodes.

File Storage can be connected to several products simultaneously, including those in different pools. For example, you can connect the storage to a dedicated server in the SPB-2 pool and a cloud server in a segment of the ru-7a pool. Both servers will be able to use the storage simultaneously.

File Storage is powered by cloud platform network volumes with triple replication of disk volumes.

You can work with File Storage in the control panel, via OpenStack CLI or Terraform.

The product supports user types and roles, projects, and project limits and quotas.

Records of File Storage operations are saved in audit logs.

Protocols

Supported protocols:

NFSv4 — for connecting the storage to servers running Linux and other Unix systems;
CIFS SMBv3 — for connecting the storage to Windows servers.

Limits and restrictions

The minimum File Storage size is 50 GB and the maximum is 50 TB.

Throughput values and read/write IOPS limits depend on the File Storage type.

File storages of the same type in different pool segments may have different limits. For example, if two File storages with the SSD Universal type are in different segments (the first in ru-1c and the second in ru-8a), their limits will differ.

You can test File Storage performance.

	HDD Basic (all available pool segments)	SSD Universal	SSD Fast
Throughput throughput (read, 4 MB blocks)	40 MB/s (for NFSv4) 100 MB/s (for CIFS SMBv3)	200 MB/s	300 MB/s
Throughput throughput (write, 4 MB blocks)	100 MB/s	200 MB/s	500 MB/s
Number of operations (read, 4 KB blocks)	320 IOPS	7000 IOPS	25000 IOPS
Number of operations (write, 4 KB blocks)	120 IOPS	4000 IOPS	15000 IOPS

File Storage types

HDD Basic — storage for large amounts of data that do not need to be frequently read or overwritten, such as backups, archives, or documents;
SSD Universal — primary storage for data if you cannot connect a disk. Also suitable for storing data sets and models in machine learning clusters and storing (raw) data for high-performance computing. Can be used in Kubernetes as a Persistent Volume with the ReadWriteMany access mode (mounting the volume to multiple nodes);
SSD Fast — storage with increased I/O performance. Also intended for storing data sets and models in machine learning clusters and for storing (raw) data for high-performance computing. Can be used in Kubernetes as a Persistent Volume with the ReadWriteMany access mode (mounting the volume to multiple nodes).

File Storage types differ in throughput values and number of read and write operations. See the Limits and restrictions table for details.

Different File Storage types are available in different pool segments. You can check type availability in the File Storage availability matrix.