CPU Scheduling & Page Replacement Simulations

Python simulations benchmarking FCFS vs SJF (CPU scheduling) and FIFO vs LRU (page replacement) across varied workloads, with matplotlib visualizations and CSV export.

Project Structure

.
├── main.py          # Entry point
├── scheduling.py    # FCFS vs SJF simulation and plotting
├── paging.py        # FIFO vs LRU simulation and plotting
├── fcfs.py          # FCFS algorithm
├── sjf.py           # SJF algorithm
├── fifo.py          # FIFO page replacement
├── lru.py           # LRU page replacement
├── process.py       # Process data class
└── results/         # CSV output (git-ignored)

Usage

pip install -r requirements.txt
python main.py

Charts are displayed inline. CSV results are saved to results/.

To run with custom parameters, call the simulation functions directly:

from scheduling import run_scheduling
from paging import run_paging

run_scheduling(num_processes=100, seed=43)
run_paging(num_seeds=10, num_frames=5, reference_length=100)

Algorithms

CPU Scheduling

Algorithm	Strategy	Preemptive
FCFS	Execute in arrival order	No
SJF	Execute shortest burst time first	No

Page Replacement

Algorithm	Strategy
FIFO	Evict the oldest loaded page
LRU	Evict the least recently used page

Results

FCFS vs SJF

SJF consistently outperforms FCFS across all dataset sizes. The convoy effect in FCFS — short processes blocked behind long ones — is the primary driver of the gap.

Processes	Waiting / Response improvement	Turnaround improvement
10	−38.4%	−29.9%
100	−36.6%	−35.9%
10,000	−29.4%	−29.4%

FIFO vs LRU

On random reference strings, both algorithms perform similarly. FIFO edges ahead slightly — random data has no temporal locality for LRU to exploit.

On structured strings with a frequently-accessed page, LRU significantly reduces page faults by keeping hot pages in memory. FIFO evicts based purely on load order and misses them repeatedly.

Key Takeaways

• SJF's advantage over FCFS shrinks as process count grows but never disappears
• LRU's advantage over FIFO is invisible on random data — workload shape is what matters
• Benchmarking OS algorithms requires realistic access patterns to surface meaningful differences

Tech Stack

• Python 3
• matplotlib — visualizations
• numpy — bar chart positioning
• csv — result export