What is Virtual Memory?

Virtual memory lets a process use more address space than the physical RAM available. Pages reside on disk and are brought in only when accessed — demand paging. The illusion of a huge address space is preserved by the MMU and the page table.

Page Table Valid/Invalid Bit

Each page-table entry has a valid bit:

• 1 → page is in memory; translation is valid.
• 0 → page is on disk (or illegal). Access triggers a page fault.

Page Fault Sequence

Performance of Demand Paging

Effective access time = (1 - p) memory_access + p page_fault_service_time, where p is the page-fault rate.

If memory access = 100 ns and page-fault service = 8 ms = 8,000,000 ns:

• p = 0.001 → EAT = 0.999100 + 0.0018,000,000 ≈ 8,099 ns. The system is 80x slower than no paging.
• p = 0.0001 → EAT ≈ 900 ns. Reducing fault rate by 10x almost recovers performance.

So a low page-fault rate (under 0.001) is essential.

Pure vs Pre-paging

• Pure demand paging — load only when faulted. Maximum laziness.
• Pre-paging — bring in expected pages proactively (e.g., on process start).

Copy-On-Write

fork() could be ruinous if it copied every page. Instead, parent and child share pages marked read-only. The first write triggers a page fault, the kernel copies that single page, and both processes continue. This is copy-on-write (COW).

Memory-Mapped Files

mmap() maps a file into the address space. Reading and writing memory then transparently reads/writes the file. Used heavily for shared libraries and inter-process shared memory.

Page Replacement Necessity

When all frames are occupied and a fault occurs, the OS must replace an existing page. The replacement choice critically affects performance — and is the topic of the next lesson.

Locality

Demand paging works because programs exhibit locality of reference:

• Temporal — recently accessed pages will be accessed again soon.
• Spatial — addresses near a recent access are likely accessed.

The working set of a process is the set of pages it has touched recently. If working sets fit in memory, fault rate is low.

Summary

• Virtual memory uses demand paging plus a valid bit to load pages on access.
• Page faults are expensive — disk is ~80,000x slower than RAM.
• Copy-on-write and memory-mapped files build on the same machinery.