• Suspension is the bridge between Memory Management and Process Management.
• Emphasize the difference between "Blocked" (waiting in RAM) and "Blocked/Suspend" (waiting on disk).
• Compaction often requires suspension: all processes are "frozen" and moved, which is why it feels like the computer has "locked up."
• Thrashing: Explain that if the OS swaps too aggressively, the disk light stays on constantly and CPU utilization drops to near zero.

• Historical Context: Overlays were widely used in the DOS era (dealing with the infamous 640KB RAM barrier) and early mainframes before Virtual Memory (Paging) became universal.
• The Key Distinction for Exams: Unlike Swapping or Virtual Memory—which are 100% transparent to the programmer and handled by the OS/Hardware—Overlays are entirely driven by the user-space software design.
• Classic Example: A two-pass compiler. Pass 1 handles lexical analysis and syntax trees. Pass 2 handles optimization and code generation. They never need to coexist in RAM simultaneously, making them perfect candidates for overlays.

To give your students a rich historical perspective, you can explain that Overlays emerged in the late 1950s and 1960s, an era when hardware memory was built using primitive "magnetic cores" (literally tiny metal donuts strung on wires). Because core memory was incredibly expensive—costing upwards of $1 to $2 per byte—computers had microscopic amounts of RAM compared to their massive physical size.

The method was championed by the most dominant tech titans of the 20th century across two distinct eras: the Mainframe Era and the Personal Computer (PC) Era.

---

### 1. The Mainframe Era (1960s – 1970s)

In this era, computers filled entire rooms, cost millions of dollars, yet frequently had only 16KB to 64KB of main memory.

• IBM (International Business Machines): IBM was the undisputed king of computing at the time. When they released the landmark System/360 mainframes in 1964, the low-end models had a tiny 16KB memory limit. To make their operating system (DOS/360) fit, IBM’s own engineers designed the OS kernel using overlays, which they called Transients. Essential hardware error routines ($A$-Transients) and file services ($B$-Transients) were manually swapped in and out of a tiny 556-byte buffer in RAM as needed.
• UNIVAC & Sperry Rand: One of IBM's primary competitors, UNIVAC, utilized sophisticated overlay systems in their EXEC I and EXEC II operating systems for the UNIVAC 1107/1108 mainframes.
• NASA & The Aerospace Industry:

Bullet list ends without a blank line; unexpected unindent.

NASA's early flight computers had strict weight and power limits, meaning very little memory. The Space Shuttle Primary Avionics System Software (PASS) famously relied heavily on meticulously programmed overlays to manage navigation, liftoff, and landing sequences within strict hardware constraints.

---

### 2. The Microcomputer/PC Era (1980s – early 1990s)

History repeated itself two decades later when personal computers emerged. Although microprocessor memory was cheaper, early PC architectures introduced a new artificial bottleneck: The 640KB Barrier.

• Microsoft and IBM (MS-DOS / PC DOS):

Bullet list ends without a blank line; unexpected unindent.

When the IBM PC was released in 1981 running Microsoft's MS-DOS, it used the Intel 8086/8088 processor. Because of how the system architecture was designed, standard user applications were strictly limited to 640KB of "Conventional Memory". As software grew more complex, companies hit a brick wall. * Lotus Development Corporation (Lotus 1-2-3): The killer app of the 1980s was Lotus 1-2-3, a massive spreadsheet program that businesses ran on IBM PCs. To allow users to build large spreadsheets without running out of the 640KB RAM, Lotus developers manually chopped their software into overlays. The core math engine stayed in memory, while graph drawing modules, printing tools, and file import functions were kept on floppy disks and loaded dynamically into an overlay buffer. * Borland (Turbo Pascal / Turbo C): Borland was famous for its programming tools. Because compilers require multiple distinct steps (Lexical Analysis $rightarrow$ Parsing $rightarrow$ Optimization $rightarrow$ Code Generation), Borland integrated Overlay Managers directly into their compilers. A programmer writing a massive program in Turbo Pascal could simply check a box, and the Borland compiler would automatically generate the overlay tree structure for them.

---

### The Extinction of Overlays

The decline of overlays is directly tied to an engineering debate at IBM in the late 1960s. An IBM researcher named David Sayre argued that automated Virtual Memory (Paging) handled by hardware and the OS could perform just as well as, or better than, a human programmer designing complex overlay structures.

By the mid-1970s for mainframes, and the mid-1990s for PCs (with the release of Windows 95 and Linux running in Intel "Protected Mode"), Virtual Memory became standard. The OS took over memory tracking entirely, relieving programmers of the massive burden of designing overlays.

> An Anecdote for Class: > You can tell your students that in the 1980s, being an "Overlay Architect" was a highly praised, highly paid specialty skill. A single mistake in tracking your code dependencies could cause a program to overwrite its own active loop, resulting in spectacular system crashes!

• Real-World Examples: The Mach kernel (which forms the core of Apple's macOS/iOS XNU hybrid), QNX (used heavily in critical automotive and medical systems due to high reliability), and seL4 (mathematically proven secure microkernel).
• The Tanenbaum-Torvalds Debate: In 1992, Andrew Tanenbaum (creator of MINIX) and Linus Torvalds had a famous debate. Tanenbaum argued Linux was obsolete because it was monolithic; Torvalds argued microkernels were impractical due to performance costs.
• Teaching Concept: Highlight that a microkernel trades sheer CPU velocity (performance) for bulletproof architectural safety and stability.