LTO Professional cheat sheet

LTO is a linear tape format developed by the LTO Consortium (Hewlett Packard Enterprise, IBM, Quantum); the format specification is licensed to media and drive manufacturers.

Recording is linear serpentine: data is written in tracks along the tape length, forward and backward, with the head stepping to the next track at each wrap; a full pass over all wraps is one data set sequence.

A data set is the fundamental written unit; each data set spans all active tracks and is protected by an error-correcting code (Reed-Solomon based) plus CRC, so a single media defect is normally recoverable.

LTO uses timing-based servo (TBS): prewritten servo bands let the drive position the head; servo format changes at LTO-8 and again at LTO-9, which is why newer drives cannot write older generations beyond a defined read-back window.

Native capacity and native transfer rate are the uncompressed figures; compressed figures assume a 2.5:1 ratio and are marketing values, not guaranteed throughput.

The drive writes a servo-format and manufacturer data set at load; a write error at the end of a partition or a full medium causes an early warning (EWE) and relocation of data.

Hard read errors are counted per drive; a drive that exceeds the error threshold on a cartridge should be suspected of head or media problems, not automatically the tape.

WORM cartridges enforce append-only writes in hardware; they cannot be reformatted or overwritten, and are used for retention and legal-hold requirements.

Tape is a sequential-access medium: random reads cost a locate plus a reposition, so throughput depends on streaming, large block sizes, and adequate host and SAN buffering.

Generation-to-native-capacity anchors: LTO-5 1.5 TB, LTO-6 2.5 TB, LTO-7 6.0 TB, LTO-8 12 TB, LTO-9 18 TB; treat later generations as roadmap items, not shipping facts, unless confirmed.

LTO-7 Type M (M8) media is LTO-7 media initialized to hold 9.0 TB, usable only in LTO-8 drives; it is a capacity step, not a new generation.

The roadmap has historically targeted roughly doubling capacity per generation with a multi-generation plan (originally to LTO-12); public roadmap statements are intentions and slip.

Each generation defines a native transfer rate; LTO-8 is about 360 MB/s and LTO-9 about 400 MB/s native, so a single drive can outrun a single HDD stream.

Backward compatibility rule of thumb: a drive reads its own generation and the previous one, and writes its own and one generation back with restrictions; check the specific drive's support matrix.

LTO-9 introduced a new servo format and 5.2 µm track pitch era changes; LTO-8 drives cannot write LTO-9 media and LTO-9 drives have limited write support for older media.

Linear Tape-Open is an open-format licensing model, so multiple vendors build drives and media to the same specification, which sustains price competition.

Capacity planning must use native capacity and account for compression ratio actually achieved; encrypted or pre-compressed data often yields near 1:1.

Compatibility is two-dimensional: drive generation versus media generation (read/write), and drive versus host interface and software (SAS, Fibre Channel, iSCSI bridging).

WORM media requires a WORM-capable drive and WORM-aware software; a WORM cartridge in a non-WORM-capable path fails at mount or at first write.

Partitioned media (LTFS) requires a partition-capable drive and formatting; a cartridge formatted for LTFS is not directly usable by legacy backup software expecting a single partition.

Library interoperability depends on the changer control path (LTO tape library interface, SCSI medium changer) and on barcode label schema matching the library's expected format.

Encryption interoperability requires the drive, library, and key manager to agree on the key management protocol (KMIP) and on whether encryption is drive-based or application-based.

Mixing generations in one library is supported but partitioning by generation improves performance and avoids mount mismatches; import/export slots are the safe path for foreign media.

Media auxiliary memory (MAM) carries application and vendor attributes; MAM is not portable across all software, so do not rely on it as the only metadata store.

Sanitization and re-use: secure erase and re-initialization behavior differ by generation and by drive vendor; confirm the erase mode before re-purposing media.

LTFS is an open specification that stores a self-describing index and XML metadata on the tape, exposing the cartridge as a file system with directories and files.

An LTFS-formatted tape is divided into two partitions: an index partition holding the latest index, and a data partition holding file content.

The index is written at unmount and periodically; an unclean unmount forces a rebuild by scanning the data partition, which is slow and should be avoided.

LTFS is single-writer: only one host may mount a cartridge read-write at a time; concurrent writers corrupt the index, so use a media manager or library-level locking.

Appending files is efficient; modifying or deleting files in place causes index churn and fragmentation, so plan write-once, read-many workloads.

LTFS is the transport layer for interchange between vendors and for object-storage tiers that front tape; it is not a backup catalog and provides no retention enforcement by itself.

Capacity and block-size tuning (for example larger block sizes and recommended mount options) materially affect streaming performance; small files degrade throughput.

LTFS-LE and vendor extensions exist but reduce portability; standard LTFS is the safest interchange format.

Drive-level encryption (LTO-4 and later) uses AES-256-GCM with the key supplied by the host or library; the drive encrypts and the key never persists on media.

Key management is the critical control: losing the key makes the cartridge unrecoverable, and exposing the key defeats encryption; use a KMIP key manager with escrow and separation of duties.

Encryption is per-cartridge and can be policy-driven by the library; unencrypted cartridges can coexist, so verify policy coverage rather than assuming it.

WORM plus encryption addresses tamper and confidentiality, but neither proves authenticity; pair with hash manifests and signed catalogs for integrity evidence.

Air-gap is the strongest ransomware control tape offers: offline or vaulted cartridges are not reachable over the network, provided the library is not left online with write access.

Tape libraries are attack surface: harden the management interface, change default credentials, segment the SAN, and restrict who can issue erase or format commands.

Data in transit on Fibre Channel can be protected with FC-SP and in-flight encryption; SAS is point-to-point and physically bounded, which limits but does not eliminate risk.

Audit requirements typically demand logging of mounts, key use, and destruction events; retain those logs separately from the tape system itself.

Active archive means data remains online and queryable through a file system or object interface while residing on tape; the tape is not a cold vault.

Typical architecture is a disk cache or object-store front end with tape as the capacity tier, driven by LTFS or a hierarchical storage manager.

Recall latency is dominated by locate and rewind, typically tens of seconds to minutes per cartridge; design workflows around bulk or prefetch recall, not random single-file access.

Policy tiers usually combine age, access frequency, and retention class; a file recalled above a threshold should be promoted back to disk.

Active archive is not a backup: it holds the primary or reference copy, so it needs its own protection, typically a second copy on separate media or site.

Metadata and search live in the front-end catalog, not on tape; losing the catalog turns the archive into unindexed tapes, so back up the catalog independently.

Throughput planning uses aggregate drive streams; a library with N drives streaming at native rate gives roughly N times single-drive throughput minus robotics contention.

Retention and legal hold must be enforced by software policy plus WORM media where required; tape alone does not expire data.

Tape's sustainability case rests on low idle power: an unmounted cartridge draws no energy, unlike spinning disk, so kWh per TB stored per year is far lower for cold data.

TCO model components: media cost per TB, drive and library amortization, floor space and power, software licensing, offsite vaulting, and migration labor across generations.

Media cost per TB is low but media has a finite life and a finite number of load/unload and pass cycles; budget replacement and migration, not just purchase.

Migration is the hidden cost: plan for a technology refresh every two to three generations, and keep at least one generation of read capability in the drive pool.

Effective cost per TB must use native capacity and achieved compression; inflated compression assumptions understate media counts and floor space.

Power and cooling savings are realized only if disk tiers actually spin down; a disk cache that never idles erases much of the tape advantage.

Re-use and recycling programs and take-back schemes reduce e-waste; tape media and cartridges have established recycling paths, so include disposal in the lifecycle.

Compare on cost per TB per year over the retention period, including migration, not on acquisition price per cartridge.