BeeGFS Features, Quotas, Mirroring, APIs, and Deployment Guidelines

BeeGFS is a high‑performance parallel file system that offers the BeeOND (BeeGFS On Demand) feature, allowing lightweight, per‑node instances to accelerate cross‑node tasks.

Quota Support

BeeGFS provides user and group quota based on space usage, requiring underlying file systems (e.g., XFS, EXT4) on storage targets to have quota enabled; metadata targets do not need quota support. The management daemon periodically collects quota information to enforce limits.

Directory Quota Tracking

For tracking directory‑level usage, the Robin Hood strategy engine can periodically scan the file system and store file and directory metadata in an SQL database, which should be placed on flash storage to avoid slowing metadata scans.

Buddy Mirroring

Since version 2012.10, BeeGFS supports metadata and file mirroring via Buddy groups, which consist of two storage targets providing primary and secondary copies. Mirroring is synchronous, and failover automatically promotes the secondary to primary if the primary becomes unavailable.

Enabling/Disabling Mirroring

Mirroring is disabled by default for new file‑system instances and can be enabled via command‑line tools after defining a Buddy group.

Supported APIs

Striping API for creating files with custom striping patterns.

Cache API for copying data between Beeond cache instances and the global BeeGFS.

Hadoop BeeGFS connector API to expose BeeGFS as an HDFS compatible file system.

System Requirements

BeeGFS client and server components run on Linux for x86, x86_64, and PowerPC/Cell architectures. Minimal hardware includes at least 4 GB RAM for storage servers, RAID configurations for SATA drives, and fast CPUs for high‑throughput networking. Metadata servers benefit from SSDs and fast CPUs due to intensive small‑file operations.

Network Support

BeeGFS works over TCP/IP, InfiniBand (OFED verbs), Omni‑Path, and RoCE, allowing mixed‑network deployments and automatic failover when primary paths fail.

Export Options

BeeGFS can be exported as SMB/CIFS by configuring a Samba server on a BeeGFS client host, or as NFS (supported since the 2014.01 release) by configuring an NFS server on a BeeGFS client.

Adoption

BeeGFS is free and widely deployed, with over 100 customers including universities, research institutions, and enterprises in finance and oil & gas. Notable TOP500 supercomputers using BeeGFS include the Loewe‑CSC cluster at Goethe University, the Vienna Scientific Cluster, and the Abel cluster at the University of Oslo.

For more information, follow the official BeeGFS channels or scan the provided QR codes.