Getting started and typical Configurations

To get started with BeeGFS, all you need is a Linux machine to download the packages from www.beegfs.io and to install them on. If you want to, you can start with only a single machine and a single drive to do some initial evaluation. In this case, the single drive would be your management, metadata and storage target at the same time. The quickstart walk-through guide will show you all the necessary steps.

Typically, the smallest reasonable production system as dedicated storage is a single machine with two SSDs in RAID1 to store the metadata and a couple of spinning disks (or SSDs) in a hardware RAID6 or software zfs RAID-z2 to store file contents. A fast network (such as 10GbE, OmniPath or InfiniBand) is beneficial but not strictly necessary. With such a simple and basic setup, it will be possible to scale in terms of capacity and/or performance by just adding disks or more machines.

Often, BeeGFS configurations for a high number of clients in cluster or enterprise environments consist of fat servers with 24 to 72 disks per server in several RAID6 groups, usually 10 or 12 drives per RAID6 group. A network, usually InfiniBand, is used to provide high throughput and low latency to the clients.

In smaller setups (and with fast network interconnects), the BeeGFS metadata service can be running on the same machines as the BeeGFS storage service. For large systems, the BeeGFS metadata service is usually running on dedicated machines, which also allows independent scaling of metadata capacity and performance.