Skip to content

CONCEPT Cited by 1 source

Tape storage (sequential access)

Definition

Tape storage is a non-volatile magnetic storage medium where data is recorded on a long strip of tape wound in a cartridge. A tape reader spins the cartridge past a fixed read/write head, reading or writing bits as the tape sweeps underneath. Tape has been in continuous commercial use since the 1950s and remains viable as an archive tier — high capacity, low $/GB, long shelf-life, and terrible random-access latency.

Dicken's framing

"A standard LTO tape cartridge has several-hundred meters of 0.5 inch wide tape. The tape has several tracks running along its length, each track being further divided up into many small cells. A single tape cartridge contains many trillions of cells."

"Even with modern tape systems, reading data that is far away on a tape can take 10s of seconds, because it may need to spin the tape by hundreds of meters to reach the desired data."

(Source: sources/2025-03-13-planetscale-io-devices-and-latency)

Why tape survives

  • $/GB is unbeatable for cold data. Tape is cheaper per byte than any live-spinning medium and has a longer shelf- life (~30+ years unrefreshed).
  • Sequential throughput is competitive. Once positioned, modern LTO-9 drives stream at ~400 MB/s uncompressed. Bulk sequential reads are not the bottleneck.
  • Air-gap durability. Cartridges not in a drive are physically disconnected from the network — immune to ransomware, firmware attacks, misconfigured ACLs.

Production-scale deployments

  • CERN: "CERN has a tape storage data warehouse with over 400 petabytes of data under management."
  • AWS Storage Gateway VTL (Virtual Tape Library): tape archiving as a cloud service.
  • Research / legal / financial / healthcare compliance archives — regulated data that must be retained for decades often lives on tape.

When tape breaks the workload

"If you read/write to a cell 'near' the read head, it's fast. If you read/write to a cell 'far' from the read head, it's slow."

"Imagine if this system was being used to load your social media feed or your email inbox. It might take 10s of seconds or even a full minute to display. This would be totally unacceptable."

Tape fails for random-access, low-latency workloads. Transactional databases, session stores, and any user-facing read path get pushed to SSDs / HDDs instead. Tape fits the shape where:

  1. Reads happen in large bulk batches, rarely.
  2. Latency (10s of seconds per seek) is acceptable.
  3. Capacity and $/GB dominate the design trade.

Tape as historical frame

The Dicken post's broader structure uses tape as the first step in the storage-latency hierarchy:

Era Medium Random access Role today
1950s+ Tape ~seconds Archive
1960s+ HDD ~1–3 ms Bulk capacity
1980s+ SSD (NAND flash) ~16 μs OLTP
2000s+ NVMe SSD ~50 μs (round-trip over PCIe) High-performance
2000s+ RAM ~100 ns Working set

Each tier survives because it wins on a different axis — concepts/storage-media-tiering.

Seen in

Last updated · 319 distilled / 1,201 read