Advise cluster FS (i.e. FS, which can be mounted on different nodes in rw) that would be able to combine storage with hundreds Terabithia.
So it is necessary that FS used the native kernel CentOS/Debian. For this reason, unfortunately not suitable GPFS and Lustre.
Looking towards the GFS, but I highly doubt that it is able to distribute the load evenly across the physical media. As far as I read, filling in the data is linear there.