Skip to content

Glossary

A named identity in the admin server that owns volumes and authenticates with an account token. Enterprise only.

A fixed-size chunk of file data stored in object storage. The block size is configured per volume (256 KiB to 8 MiB, default 4 MiB).

The storage protocol used to access a block store. Supported values: s3, gcs, azure, oci.

A configured object storage location (bucket + optional prefix) where flexFS writes block data. Each block store is associated with a provider and region.

An estimated monthly storage cost in US dollars ($/month), computed per file as (effective_size / 1 GiB) * rate. The effective size is the smaller of the file’s logical size and its allocated block size (protecting sparse files from overcharging). The rate is a per-GiB monthly dollar amount determined by the file’s size bin — an access-age tier. Rate bins are configured per account and allow tiered pricing by data temperature. Reported by analyze.flexfs and find.flexfs.

An industry standard for exposing storage to container orchestrators. csi.flexfs implements CSI to provide flexFS volumes to Kubernetes pods.

A directory entry — a name-to-inode mapping within a directory. Dentries are stored in the metadata server.

A Linux kernel interface that allows filesystem implementations to run in user space. FlexFS uses FUSE3 to present volumes as local filesystems.

A metadata record describing a file, directory, symlink, or special file. Contains attributes such as size, ownership, permissions, and timestamps.

A registered metadata server instance in the admin server’s configuration. Linked to volumes to route mount clients to the correct metadata server.

The subdirectory within a volume that a volume token restricts access to. When set, the mount client sees only the subtree below that path.

A cloud or custom infrastructure provider (e.g., aws, gcp, azure, oci). Providers are registered in the admin server and used to organize regions and block stores.

A set of one or more proxy servers that cache block data for a region. Mount clients select the lowest-latency group using RTT probing and distribute blocks across group members using rendezvous hashing. Enterprise only.

A geographic or logical zone within a provider (e.g., us-east-1). Regions group block stores and proxy groups by location.

A consistent hashing algorithm used by mount clients to distribute blocks across proxy servers within a group. Each block is deterministically assigned to a specific proxy without requiring a central lookup.

The duration for which deleted data (both metadata and block data) is preserved, enabling time-travel mounting. Configured per volume; -1 means forever. Default is 7 days.

An identifier for the end-to-end encryption key material registered with the admin server. Used to verify that mount clients provide the correct secret when mounting encrypted volumes.

A mapping from size bin number to a per-GiB monthly dollar rate ($/GiB/month), configured per account. Rate bins enable tiered storage pricing: recently-accessed data can be priced differently from cold data. The cost computation walks backward from a file’s size bin to find the applicable rate. Used by analyze.flexfs, find.flexfs, and the statistics server for billing.

A bin that classifies files by how long since they were last accessed (atime). Bin 0 means accessed within the past 30 days, bin 1 means 30-60 days ago, and so on in 30-day increments up to bin 74 (6+ years). Used together with rate bins to compute per-file storage cost, and for usage analysis.

A logical filesystem managed by flexFS. Each volume has its own namespace, block store, metadata store, and configuration (block size, compression, end-to-end encryption, retention).

A UUID that grants access to a specific volume. Volume tokens can be scoped to a mount path (subdirectory) and carry per-token mount flags. Mount clients authenticate using a volume token.

A POSIX.1e access control list that extends standard Unix permissions (owner/group/other) with per-user and per-group entries. Stored as extended attributes (system.posix_acl_access and system.posix_acl_default). Enabled with the --acl mount flag.

A name-value pair associated with an inode, extending the standard POSIX attribute set. Used by ACLs, SELinux labels, and user-defined metadata.