best practice for many storage daemons

[emme]

good afternoon,

  is there a best practice/guide/ suggestion for building an infrastructure based on bareos for backing up many servers and clients (500+)?

i was confused if it is better to have a server for director deamon only, then add many servers for only storage deamons, and in this case to have big local disks on each storage server, or builing a common scale out ceph/s3 or nfs repository, and let all storage daemons point to it via network...

thank you.