---
title: "The 1959 Manchester Atlas was the first operating system with virtual memory — called a 'one-level store' — and the idea was contested for a decade"
type: "claim"
status: "seedling"
audit_status: "capture-verified\n2026-07-12 cross-model audit (auditor: claude-fable-5): source PDF re-fetched, quotes verified verbatim. Corrected overlay-time qualifier (\"on large problems\" → source's \"planning overlay sequences\"). Appended dating caveat: Denning's \"1959\" is his dating; primary record puts commissioning at 7 Dec 1962 and the Kilburn et al. one-level store paper at 1962 (IEEE Milestone spans 1957–1962). Priority claim (Atlas first) stands.\n"
writer_model: "claude-sonnet-5"
source_url: "https://denninginstitute.com/pjd/PUBS/ENC/CRC-vm-2013.pdf"
source_author: "Peter J. Denning"
source_date: "2013"
source_quote: "The first operating system with virtual memory was the 1959 Manchester Atlas… They called their virtual memory a 'one-level store.'"
source_tier: 2
provenance: "Promotion from 10-inbox/raw/2026-07-11-hop-os-virtual-memory-in-llm-serving.md, 2026-07-12"
origin: "batch"
derived_from: "10-inbox/raw/2026-07-11-hop-os-virtual-memory-in-llm-serving.md"
date_created: "2026-07-12T00:00:00.000Z"
tags: ["virtual-memory","operating-systems","history-of-computing","manchester-atlas","denning"]
related_notes: ["claim-pagedattention-borrows-os-virtual-memory-paging","claim-denning-1968-working-set-model-solves-thrashing","claim-memory-wall-named-1994-wulf-mckee"]
audits: ["2026-07-12 claude-fable-5"]
drafted_in: ["1968-had-no-word","2026-07-13-1968-had-no-word"]
---


Per Peter J. Denning's encyclopedia account of virtual memory: "The first operating system with virtual memory was the 1959 Manchester Atlas… They called their virtual memory a 'one-level store.'" Before automatic virtual memory, programmers hand-managed overlays to fit programs into scarce physical memory — work Denning estimates once consumed "half to two-thirds" of most programmers' time, spent planning overlay sequences.

**Dating caveat (audit 2026-07-12):** "1959" is Denning's dating and tracks the design, not the running machine. Atlas's detailed design was completed at the end of 1959, but the machine was commissioned on 7 December 1962, and the team's own primary paper — Kilburn, Edwards, Lanigan & Sumner, "One-Level Storage System," *IRE Transactions on Electronic Computers* EC-11 (1962), which Denning himself cites alongside Fotheringham (1961) — appeared in 1962. The IEEE Milestone accordingly dates the invention as a span: "Atlas Computer and the Invention of Virtual Memory 1957–1962." The priority claim (Atlas first) is uncontested; the single-year date is looser than a bare "1959" suggests.

The idea was not received as an obvious win. Denning records that virtual memory was "the subject of intense controversy after its introduction in 1959," debated for roughly a decade until **David Sayre**'s 1969 work settled the argument (by demonstrating that automatic, OS-managed memory replacement could match or beat hand-tuned manual overlays).

This is the origin point of the address-mapping principle that [[claim-pagedattention-borrows-os-virtual-memory-paging|vLLM's PagedAttention]] re-imports, 64 years later, to manage the LLM KV cache. It also seeded a second, related unresolved-performance problem — multiprogramming pushing the same scarce-fast/abundant-slow memory hierarchy into collapse — addressed by [[claim-denning-1968-working-set-model-solves-thrashing|Denning's 1968 working-set model]].

> [!note] Seek's commentary:
> The genuinely surprising fact here isn't the 1959 date, it's the decade of controversy. Virtual memory reads today as an obviously-correct abstraction; it wasn't obvious in real time, and it took a specific empirical result (Sayre 1969) rather than an a priori design argument to close the debate. Worth remembering the next time an architectural choice looks "obviously correct" in the LLM-serving literature.
> — Seek