Sizing a Long-Term Tape Archive in Petabytes
A backup administrator rarely thinks in petabytes on day one. You think in terabytes: the file server, last night's incremental, one LTO cartridge. Petabytes arrive later, when someone asks how many slots the library needs in year seven and you finally add up everything the retention policy has been telling you to keep.
Type the terabytes your backup chain accumulates and read the petabyte figure you will have to justify to whoever signs off on slots, floor space and off-site vaulting.
Value in PB = Value in TB × 0.001 — both are decimal SI units, so it is a straight move of the decimal point three places. A chain that writes 953 TB in its first year is 953 × 0.001 = 0.953 PB.What Fills the Vault
One Cartridge Is 0.018 PB
The Chain, Not the Dataset
Copies Multiply Everything
Slots Are the Real Budget
Turning a Backup Chain into a Vault Total
Work from what the backup catalogue reports, not from the size of the production volumes. Your reports know how many terabytes actually left the staging area each month, and that is the number the archive has to swallow.
Total One Year of Writes
Add the full backups and the incrementals for twelve months. For a 50 TB source with monthly fulls and a 1 TB average daily incremental that is (12 × 50) + (353 × 1) = 953 TB.
Multiply by the Retention Term
Seven years of that chain is 6,671 TB. Type it into the TB field and the PB figure appears as you type — 6.671 PB — without pressing anything.
Apply Your Copy Count and Record It
Double the result for a second vault copy, then use the copy button beside either field to lift the bare number into the capacity plan — no unit text to clean up.
Work Backwards from a Budget
Both fields are editable and the swap button reverses the pair. If the library you can afford tops out at 2 PB, put 2 into the PB side and read the 2,000 TB your policy may generate.
Commas and dots are both accepted as the decimal mark and spaces in the input are ignored, so a figure pasted from a European capacity report converts without editing.
Backup Chain Growth Across a Retention Term
The table follows one modest environment — a 50 TB source, one full backup per month, daily incrementals averaging 1 TB — through the retention terms compliance usually argues about. Cartridge counts assume LTO-9 at its native 18 TB, no compression, rounded up to whole cartridges.
Retention Term to Vault Petabytes
| Retention | Data written (TB) | One vault copy (PB) | LTO-9 cartridges | Two copies (PB) |
|---|---|---|---|---|
| 1 year | 953 TB | 0.953 PB | 53 | 1.906 PB |
| 3 years | 2,859 TB | 2.859 PB | 159 | 5.718 PB |
| 5 years | 4,765 TB | 4.765 PB | 265 | 9.53 PB |
| 7 years | 6,671 TB | 6.671 PB | 371 | 13.342 PB |
| 10 years | 9,530 TB | 9.53 PB | 530 | 19.06 PB |
Two things jump out. A source that never grows still pushes a ten-year archive close to 10 PB, purely through repetition — and the cartridge count crosses several hundred long before the petabyte figure looks alarming, which is why libraries run out of slots before they run out of raw capacity.
Cartridge-Count Sanity Check
Convert the chain to PB, then divide by native cartridge capacity — the fastest way to tell whether a proposed library has slots for the whole retention term or only for year one.
Retention Math Without a Spreadsheet
Retype the terabytes for a different retention term and the petabyte total re-renders live — useful when someone across the table asks "and if we kept it for ten years?"
Every Unit the Backup Report Uses
The searchable dropdown on each side lists the whole data-storage family, so a job log quoted in GB or a throughput figure in MB/s can be dropped into the same two fields.
Readable at Archive Scale
Results carry up to eight decimals with spaced thousands and switch to scientific notation past ten billion, so a byte-level total from a catalogue export stays legible.
Tape Archive Questions
Why does the vendor quote 45 TB on an 18 TB cartridge?
Because the larger figure assumes 2.5:1 hardware compression. Every generation since LTO-6 is marketed that way: LTO-7 is 6 TB native and 15 TB compressed, LTO-8 is 12 TB and 30 TB, LTO-9 is 18 TB and 45 TB. Video, images and encrypted volumes barely compress at all, so size the vault on native capacity.
Does a full-plus-incremental chain really grow that fast over a decade?
It does, because you are storing a picture of the data every month rather than the data once. Each monthly full re-writes the whole dataset whether or not it changed, so a static 50 TB source reaches 9.53 PB across ten years. Expiring sets on schedule and using synthetic fulls cuts that sharply.
How much does the 3-2-1 rule add to the petabyte total?
Usually it doubles the tape half of the plan. Three copies on two media types with one off site typically means production data, a disk staging copy, and two tape copies — one in the library, one in a vault across town. A 6.671 PB seven-year chain becomes 13.342 PB of cartridges to buy and track.
How long will the cartridges stay readable before I have to migrate them?
Vendors rate LTO media for around 30 years in a cool, stable room, but the drives set the schedule. Since LTO-8 a drive reads and writes only one generation back, so an LTO-9 unit handles LTO-8 media and nothing older. Plan a migration pass every five to seven years and budget drive hours: it re-reads the entire archive.
How long does a petabyte-scale restore actually take?
An LTO-9 drive streams about 400 MB/s of native data, roughly 42 minutes per terabyte. A petabyte on one drive is about 694 hours — near enough 29 days of continuous streaming, or under four days across eight drives. Add cartridge exchange and locate time and scattered small-file restores run far slower still.
No comments yet. Be the first to comment!