Thrashing and Working Set Model

Thrashing and the Working Set Model are key concepts in Operating Systems that deal with memory management and system performance.

Thrashing

6

Thrashing occurs when a system spends more time swapping pages between main memory and disk than executing actual processes.

Why it happens:

The system is overloaded with too many processes.

Each process doesn’t have enough frames (memory pages).

Frequent page faults occur → OS keeps swapping pages in and out.

Effects:

CPU utilization drops sharply.

Disk activity (I/O) increases dramatically.

System performance becomes extremely slow.

Simple intuition:

Instead of doing useful work, the system is “busy moving data around.”

Working Set Model

6

The Working Set Model is a strategy to prevent thrashing by ensuring each process has enough memory.

Definition:

The working set of a process is:

The set of pages that the process is actively using during a given time window (Δ).

Key idea:

Programs exhibit Locality of Reference:

They tend to reuse a small subset of pages at a time.

If all these pages are in memory → few page faults.

If not → thrashing risk.

How it works:

OS tracks pages used in the last Δ time units.

Allocates enough frames to hold this working set.

If memory is insufficient:

Reduce number of processes (load control).

Relationship Between Them

Thrashing = Problem

Working Set Model = Solution

Aspect	Thrashing	Working Set Model
Nature	Performance issue	Memory management technique
Cause	Too many page faults	Prevents excessive page faults
Goal	Avoid it	Ensure sufficient frames per process
Result	Low CPU utilization	Stable performance

Quick Analogy

Imagine studying for exams:

Thrashing: Constantly switching between books without focusing → no progress.

Working Set: Keeping only the necessary books on your desk → efficient study.

Memory Management in Linux:

Linux memory management is the mechanism by which the operating system efficiently allocates, tracks, protects, and reclaims memory for processes and the kernel.

Goals of Memory Management

Linux memory management aims to:

Maximize CPU utilization

Provide process isolation and protection

Support multitasking

Use RAM efficiently

Minimize disk I/O

Enable virtual memory

Types of Memory in Linux

1. Physical Memory

Actual RAM installed in the computer.

2. Virtual Memory

Each process gets its own virtual address space.

Advantages

Process isolation

Larger address space than physical RAM

Easier programming model

Virtual Memory System

Linux uses paging.

A process’s virtual memory is divided into pages (commonly 4 KB).

Physical memory is divided into frames.

Mapping:

Virtual page → Physical frame

This mapping is handled using page tables.

Address Translation

CPU-generated virtual addresses are translated to physical addresses using:

Page Table

MMU (Memory Management Unit)

TLB (Translation Lookaside Buffer)

Linux Process Memory Layout

Typical process memory layout:

+-------------------+
| Stack             |
|        ↓          |
+-------------------+
| Shared Libraries  |
+-------------------+
| Heap              |
|        ↑          |
+-------------------+
| Data Segment      |
+-------------------+
| Text Segment      |
+-------------------+

Segments Explained

Text Segment

Executable code

Usually read-only

Data Segment

Global/static variables

Heap

Dynamic memory (malloc, new)

Grows upward

Stack

Function calls/local variables

Grows downward

Paging in Linux

Linux uses:

Demand Paging

Copy-on-Write (COW)

Swapping

Demand Paging

Pages are loaded into RAM only when needed.

If page not in memory:

Page fault occurs

Kernel loads page from disk

Benefits:

Reduced memory usage

Faster process startup

Copy-on-Write (COW)

Used during fork().

Parent and child initially share pages.

Actual copying happens only when one modifies a page.

Benefits:

Saves memory

Improves performance

Swapping

When RAM is insufficient:

Inactive pages moved to swap space on disk.

Swap can be:

Swap partition

Swap file

Too much swapping may cause: