Skip to main content

Memory Management

The goal of memory management is to provide each process with its own private address space and maximize the utilization of physical RAM.

Logical vs. Physical Address​

  • Logical Address: Generated by the CPU. Programs use these addresses.
  • Physical Address: The actual location in the RAM hardware.

The MMU (Memory Management Unit) is a hardware component that translates logical addresses into physical addresses on the fly.

Address Translation Techniques​

Segmentation​

Memory is divided into segments of variable size (e.g., Code, Data, Stack).

  • Pro: Matches the programmer's view of a program.
  • Con: Causes External Fragmentation (free memory in small, non-contiguous chunks).

Paging​

Physical memory is divided into fixed-size blocks called Frames. Logical memory is divided into blocks of the same size called Pages.

  • Pro: Eliminates external fragmentation; allows non-contiguous memory allocation.
  • Con: Causes Internal Fragmentation (unused space within the last page/frame).

Page Table​

A data structure in the kernel used by the MMU to store the mapping from Page Number to Frame Number.

  • TLB (Translation Lookaside Buffer): A hardware cache that stores the most recently used page-to-frame translations to speed up memory access.

Virtual Memory​

Virtual memory allows a process to execute even if only a part of it is in physical memory.

  • Demand Paging: Loading a page into memory only when it's accessed.
  • Page Fault: A hardware exception triggered when a process accesses a page not currently in memory. The kernel then fetches it from disk (the swap area).

Page Replacement Algorithms​

When memory is full and a new page needs to be loaded, the OS must choose a page to evict.

  • FIFO (First-In-First-Out): Simplest but prone to Belady's Anomaly (more frames can lead to more page faults).
  • LRU (Least Recently Used): Evicts the page that hasn't been accessed for the longest time. Generally efficient but hard to implement perfectly in hardware.
  • Clock (Second-Chance): A practical approximation of LRU. Uses an "access bit" in the page table entry.

Memory Allocation in Kernel​

The kernel itself needs to allocate memory dynamically for its own structures.

Buddy System​

Splits a large block of memory into powers of 2. When a block is freed, it's merged with its "buddy" if available.

  • Pros: Fast allocation and deallocation; reduces external fragmentation.

Slab Allocator​

Pre-allocates caches for frequently used kernel objects (e.g., PCBs, inodes).

  • Pros: Zero internal fragmentation for specific object sizes; extremely high performance.