It's nothing really exotic. It's a "high density" storage system from RAID, Inc., with 84 drives in a 5U form factor. We got 12TB drives, so raw capacity is 1008TB. It's connected to a Dell R640 server running CentOS 7. I'll configure it with RAID 6, with a some hot spares. I'll also buy a few cold spares. The file systems on it will be shared via NFS to various client systems. We have a rack of compute servers for data analysis which are connected to the file servers through a 10Gbit switch.

Just wondering - why RAID6 and not ZFS?

I don't have a lot of experience with ZFS and I've been reluctant to use it on Linux, due to the licensing issues. But I am interested in ZFS and will probably look into it at some point. I'd probably need a different storage solution, one that presents all the drives individually to the host, since that's what ZFS wants.

Cool project! ZoL is rock solid and it's easy to setup. You could switch it over in a weekend. The standard HBA last time I bought one was the M1015 flashed into IT mode. ZFS data integrity features alone are worth it for that much data. If you mention what you are up to on freenode#zfsonlinux people will jump at the opportunity to help.

Any chance at bringing someone new in to take over the brunt of the design and implementation? It sounds like a natural point to transition.

Isn't RAID6 iffy re bit rot at PB levels?