An introduction to GlusterFS

1) GlusterFS was a CA-system (in the CAP-theorem way), which means that availability is favored over consistency (i.e. split-brains are possible). In ceph, which is a CP-system, split-brains are impossible, but the cluster is considered down, if consensus (majority) cannot be reached

2) Last I checked, it was the *clients* job to do all copies, and "patrolling" is only done when accessing data (i.e. silent bitrot is possible). Ceph on the other hand has the duplication (and re-duplication in case of trouble) as part of the servers. Also in ceph, patrol-reads are a standard feature (and done server-side).

I thus strongly prefer Ceph, even if I have not affiliation to either..

Please correct me, if my understanding is wrong...