Skip to main content

Core concepts

File Storage is a storage service that allows multiple servers to access the same file system over a network. This document explains the core concepts, states, performance characteristics, operational recommendations, and IAM roles for using File Storage.

File system​

A File Storage file system is the primary resource that manages files and directories in a hierarchical structure. You can read, write, and delete data by file, and access it over the network using standard protocols such as NFS or SMB.

File system types​

File Storage provides the following two types depending on the use case.

File system typeDescription
Infinite file systemScalable file system
- Provides SMB and NFS file services
- Provides multiple shared volumes in one file system
- Can be used without a fixed storage size limit
Basic file systemGeneral-purpose NFS file system
- Supports NFS protocol
- Can be created from 1 TiB to 16 TiB in 0.5 TiB increments

File system lifecycle and state​

A file system starts in the Creating state and is configured in the specified network environment. When configuration is complete, it transitions to Active, and clients can mount and use it.

Deleting indicates that file system deletion is in progress. After resources are released, the file system is removed from the list and billing stops.

StateDescription
CreatingCreating the file system
ActiveFile system is running normally and available
UpdatingFile system configuration operation is in progress
DeletingDeleting the file system
ErrorAn error occurred in the file system

File system performance​

File Storage is optimized for environments where multiple servers access files at the same time to share data. Because it is network-based storage, actual performance is affected significantly by client specifications and network settings.

Measuring and understanding File Storage performance​

Cloud file storage uses a structure where multiple clients share resources, so performance characteristics between the client and file storage must be considered together. KakaoCloud provides performance guidance based on standard I/O testing with fio.

The following values measure ranges close to actual user-perceived performance and can vary significantly depending on VM specifications and network settings.

Key variables

  • VM network bandwidth
  • NFS mount option, such as nconnect
  • Number of concurrent jobs

Client reference

ItemSpecification
VM flavor8 vCPU / 16 GB RAM
Network10 Gbps
OSLinux
ProtocolNFS v4.1
Mount optionnconnect=1 / 4 / 16
Write modeAsync

Performance results

Test itemnconnectI/O patternBlock sizeIOPS
Random Readnconnect = 1100% Read4 KBup to 30,000
Random Readnconnect = 4100% Read4 KBup to 57,000
Random Readnconnect = 16100% Read4 KBup to 66,000
Random Writenconnect = 1100% Write4 KBup to 28,000
Random Writenconnect = 4100% Write4 KBup to 42,000
Random Writenconnect = 16100% Write4 KBup to 47,000

Increasing the nconnect value increases the number of parallel connections and can improve IOPS.

File system operational recommendations​

Refer to the following recommendations to operate File Storage reliably and reduce performance degradation.

Store large files​

There is no limit on the size of a single file, but operational efficiency and stability should be considered. Single files of several TB or more can increase retry costs during transfer and lengthen recovery time after failures. For large data, splitting it into multiple files is recommended.

Limit metadata-heavy operations​

Creating or deleting a large number of files in a short time can increase system load. Recursive commands such as chmod -R, chown -R, and ls -R from a top-level directory can affect overall performance and should be used carefully.

Use symbolic links for file references. Hard links can cause internal lock contention and degrade overall response performance.

Manage the number of files in a directory​

Storing an excessive number of files in one directory can cause performance degradation due to internal distributed processing. If the number of files continues to increase, separate directories by date (YYYY/MM/DD) or category.

warning
  • Recommended: Fewer than 50,000 files per directory
  • Caution: File creation errors can occur with 500,000 or more files

Limit concurrently accessed files and workloads​

File Storage optimizes performance by caching metadata in memory. If requests continuously exceed the cache range, overall performance can degrade and affect other users' access. Operate workloads within the recommended range.

warning
  • Recommended: Fewer than 1 million managed files in total
  • Client: Up to 500,000 concurrently open files per mount point

IAM-based role management​

File Storage follows IAM role-based access control (RBAC). Permissions to view, create, modify, and delete file systems and backups are granted based on IAM roles. For how to manage user roles, see Manage user IAM roles.

PermissionProject AdminProject MemberProject ReaderFile Storage ManagerFile Storage Viewer
View file system✓✓✓✓✓
Create, edit, delete file system✓✓✓