Why Modern Operating Systems Need Virtual Memory: A Deep Dive

CPU and main memory are scarce resources shared by all processes; the OS uses a CPU scheduler and a virtual memory system to manage them. Before discussing why virtual memory is needed, it is essential to understand that virtual memory acts as an intermediate layer between physical memory and processes, hiding the physical address space and providing a simpler, more powerful interface.

Figure 1 – Intermediate layer between processes and the OS

If an OS were designed from scratch, allowing processes to access physical addresses directly would seem natural; early OSes did exactly that. Modern OSes, however, introduce virtual memory: processes use virtual addresses that the Memory Management Unit (MMU) translates into physical addresses before accessing memory.

Figure 2 – Virtual memory system

Main memory, though limited, offers random‑access speeds up to 100,000× faster than disk, making efficient use of its speed crucial for program performance.

The OS manages memory in pages; when a process accesses data not present in RAM, the OS loads the required page. Virtual memory provides three key benefits:

It uses disk as a cache, speeding up access to frequently used memory regions.

It gives each process an independent address space, simplifying linking, loading, and enabling shared libraries.

It controls process access to physical memory, isolating processes and enhancing system security.

Cache

Virtual memory can be viewed as a disk region; the frequently accessed portion is cached in RAM as pages, accelerating CPU access. Disk provides large storage, while RAM offers fast access, creating the illusion of a large, fast memory pool.

Figure 3 – Virtual memory, main memory, and disk

Virtual pages can be in three states: unallocated, uncached (only on disk), or cached (present in RAM). When a process accesses an uncached page, a page‑fault interrupt occurs. If the page table lacks a mapping, the OS creates one; otherwise, the OS loads the page from disk into RAM.

Figure 4 – Page fault handling

When RAM is full, the OS evicts suitable physical pages back to disk, a process called page replacement. Page faults and replacement are part of the paging algorithm, which aims to use disk as a cache for RAM to improve program efficiency.

Memory Management

Virtual memory gives each running process an isolated address space; on a 64‑bit OS each process can have up to 256 TiB, with kernel and user spaces each occupying 128 TiB. Some OSes use 57‑bit virtual addresses to provide 128 PiB of addressable space. Because each process’s virtual space is independent, they can use the full range from 0x0000000000000000 to 0x00007FFFFFFFFFFFF.

Figure 5 – Virtual memory space layout

The OS implements address translation via multi‑level page tables. Linux introduced a four‑level page table in kernel 2.6.10 and a five‑level table in 4.11, with further extensions possible for larger address spaces.

Figure 6 – Four‑level page table structure

In this hierarchy, the lowest 12 bits are the page offset; the remaining bits are split into four groups, each indexing a level of the page table to locate the corresponding physical page.

Because page tables map virtual pages to physical pages, multiple processes can share the same physical memory. The fork system call exemplifies this: the child process receives a copy of the parent’s page tables, not the physical pages themselves, allowing both processes to reference the same memory.

Figure 7 – Shared memory between parent and child after fork

Memory Protection

User programs must not modify read‑only code segments, nor access kernel memory or other processes’ private data. Virtual memory’s page tables store access rights (read, write, execute) for each mapping, enabling the MMU to enforce these permissions and reduce security risks.

Figure 8 – Read, write, and execute permissions

Conclusion

Virtual memory combines disk and RAM advantages, presenting a fast, large‑capacity storage illusion to processes. It isolates process address spaces, uses multi‑level page tables for translation, enables memory sharing to reduce overhead, and enforces protection to improve security. Understanding these design principles offers valuable insights for broader software engineering challenges.

Key reasons why Linux needs virtual memory:

It merges disk and RAM to provide seemingly fast and large storage for processes.

It gives each process an independent address space with multi‑level page tables, allowing shared physical memory and simplifying linking, loading, and allocation.

It controls process access to physical memory, isolating processes and enhancing system security.