Digital Garden

Recent Notes

BeeGFS course overview

1 min read

BeeGFS

  • Designed for performance-critical environments.

  • Distributed parallel file system architecture.

  • Supports integration with common Linux workflows.

  • Graphical tools are primarily used for monitoring.

Architecture

Management

  • Rendezvous point for new servers and clients.
  • Not data-path performance critical.
  • Stores cluster metadata, not user payload data.

Storage

  • Stores chunk files.
  • Can use multiple storage targets (RAID, NVMe, SSD, HDD).

Metadata

  • stores location of the chunk files
  • metadata server can only have one target
  • can be multiplied

Client

  • mounts the fs
    • default /mnt/beegfs

Optional

  • Monitoring
    • Admon
    • Grafana

BeeGFS course day 1

Striping

  • Chunk size: 64 KB (least) to 32 MB (or higher for really large files)
  • change targets and chunks size per directory
    • lees targets and chunksize for smaller files
    • more for larger files
    • separate smaller and larger files
  • set via beegfs-ctl

Multi-mode

  • Common (check presentation for setup)
    • register at mgmt server
      • -s for unique id AND -S unique name
      • Think of a system for this

See also: BeeGFS course overview, BeeGFS course day 2, MOC Computing DevOps and Software

Link to original

See also: BeeGFS course day 2, parallel file system performance notes, MOC Computing DevOps and Software

Graph View

Backlinks