Workload summary
- Total usable data: ~500 TB across six servers.
- Mostly-static archive set: media files that seldom change (bulk of 500 TB).
- Frequently-changing set: ~20 TB changes often; needs daily archival and multi-tier rotation (weekly/monthly/quarterly/annual).
- Restore testing: quarterly restore test for every tape.
- Cost constraints: used gear is acceptable.
Recommended topology (3-2-1)
- Primary: live data on the six servers.
- Secondary (disk, on-site): a backup server/NAS sized to hold daily versions of the ~20 TB active set (snapshots/dedupe), enabling fast restores.
- Tertiary (tape, offline/off-site capable): LTO library + media for long retention and ransomware-resistant copy.
Why disk-first for the daily-changing set: daily versioning to tape is possible but often operationally expensive.
A common, cost-effective pattern is D2D2T: daily to disk, weekly/monthly/quarterly/annual to tape.
LTO-8 vs LTO-10 (fit for this workload)
| Category | LTO-8 | LTO-10 |
|---|
| Native per tape |
12 TB (see lto.org LTO-8) |
30 TB (and newer 40 TB cartridge spec) (see lto.org LTO-10) |
| Tapes for ~500 TB archive |
~42 tapes (+spares) |
~17 tapes at 30 TB (or ~13 at 40 TB) |
| Weekly full for 20 TB set |
~2 tapes per weekly full |
~1 tape per weekly full (30 TB) |
| Used-market economics |
Usually best value (drives + media more available used) |
Newest gen often carries a premium; used supply may be thinner |
| Upgrade/migration risk |
Good compatibility options in mixed environments |
LTO-10 drives only use LTO-10 media (plan migrations carefully). See compatibility chart. |
| Quarterly verify “every tape” |
More tapes to read, but cheaper to own enough slots/media |
Fewer tapes to manage/verify, but higher entry cost |
What I would buy (cost-effective, used-friendly)
- Used 48-slot SAS library + 1 LTO-8 SAS drive (48 slots is a practical minimum once you factor archive + GFS retention + spares).
- Barcode labels and at least one cleaning cartridge.
- Backup server with enough disk to hold daily versions of the ~20 TB changing set.
When LTO-10 becomes the better choice
- You strongly want to minimize tape count (operator time/handling).
- You can afford the higher drive/library cost and media premium.
- You accept the compatibility break and have a clear migration story.
Restore-test scheduling idea
If you have ~42 archive tapes (LTO-8) plus rotation sets, “every tape every quarter” is realistic if automated:
- Run one full-tape read verify per night (or per weekend), continuously.
- Track results per tape ID; retire tapes showing rising soft error rates.
Next inputs needed to size the exact tape pool: your daily change rate within the 20 TB set, and your exact GFS retention policy (how many weeklies, monthlies, quarterlies, annuals you keep).