Assuming you have LTO-5 tape, what else do you need to make LTFS work?
Let me start with a quick clarification. The relationship between the Linear Tape File System and LTO-5 is a function of the ability of LTO-5 to partition its media. Partition 0 holds directory structures and pointers that can let the tape drive seek the specific data being sought from the tape more quickly; the data itself is stored in Partition 1. Non-LTO tape technologies, such as some of the proprietary “enterprise” tape media offered by non-LTO vendors, could also adapt the Linear Tape File System to their partitioning schemes.
That said, setting up a rig using LTFS and a LTO-5 library requires a cobble of a generic server head (running Linux, Mac OS, or Windows -- all of which have, or will shortly have, a compatible LTFS driver), NFS/CIFS/SMB software if you want to expose the data store to users, and some disk for caching recently accessed files.
The guy who wrote the book on this is Rob Sims at Crossroads Systems, who did the earliest designs I saw and who has created an elegantly integrated “Tape NAS” head based on his StrongBox product. You can also build one using IBM SONAS, with its Tivoli Storage Management software providing some necessary management and integration hooks with a backend library. I have little doubt that other products will emerge in the market as this meme gains adherents.
I like this tape NAS concept a lot. We will probably need it sooner -- rather than later -- given the failure of most IT departments to do anything about managing data itself, which has resulted in a highly inefficient storage junk drawer in most shops I visit today.
This was first published in March 2012