Ceph pg

[Arthur777]

hello, i'm currently studying ceph and there's something i'm having trouble understanding: what's the real benefit of pg? I've seen examples where it was explained that in the case of a triple disk failure there was less data loss with pg than if the objects were stored directly in 3 osd but I can't understand what this changes concretely.

Thanks in advance.

 

[UdoB]

A Placement Group is one core element of Ceph to establish the flexibility and reliability everyone wants to have ;-)

There are MONs, MGRs, OSDs, MDSs and some more parts. Ceph ist complex! This should not be surprising as it if offers a lot of functions and nearly infinite scalability.

Each and every part of this architecture consumes resources of the hardware it is running on. Ceph is not really meant to be used on small computers with only two SATA-disks, a few GB of Ram and slow (1 GBit/s) not-separated networks.

That said...: beside my dayjob I run a homelab. Mainly with ZFS (which works great and is more suitable), but additionally with a sub-minimal Ceph setup. Because I can, not because it is recommended. It works. It is slow, but it works

Skim over the reference docs, their Intro talks about those basic elements: https://docs.ceph.com/en/latest/start/intro/ https://docs.ceph.com/en/latest/start/ and another chapters goes into details: https://docs.ceph.com/en/latest/architecture/

Best regards

Edit: the first link mutated...