FACULTÉ DES SCIENCES

DÉPARTEMENT D’INFORMATIQUE

INFO-F-404: Real-Time Operating Systems

Uniprocessor Scheduling Project
Joël Goossens Yannick Molinghen

1. Introduction
In this project, we consider synchronous task sets with constrained deadlines. This project consists in
• the implementation of three priority assignment algorithms (DM, EDF and Round Robin),
• a scheduler simulator,
• a report that compares the performance of DM, EDF and RR.

2. Project details
This project should be realised by group of two. There are two deliverables: the code itself preferably
written in Rust or in Python¹ and a PDF report.

2.1. Task set file

Your program has to parse task set files. A task set file is a file that contains a description of a set of
tasks with their respective offset 𝑂𝑖 , computation time 𝐶𝑖 , deadline 𝐷𝑖 and period 𝑇𝑖 . Task set files
are simple Comma Separated Value (CSV) files where the line 𝑖 encodes the 𝑂𝑖 , 𝐶𝑖 , 𝐷𝑖 , 𝑇𝑖 of task 𝑖.
An example is shown here below where 𝜏1 = {𝑂1 = 0, 𝐶1 = 2, 𝐷1 = 40, 𝑇1 = 50}, and 𝜏2 = {𝑂2 =
0, 𝐶2 = 80, 𝐷2 = 200, 𝑇2 = 200}.

0, 20, 40, 50
0, 80, 200, 200

Note that the offset of each task is given in task set files even though it is systematically 0 in the scope
of this project.

2.2. Scheduling algorithms

To determine the priority of each task, you have to implement Round Robin, Earliest Deadline First
and Deadline Monotonic algorithms. Note that
• DM assigns priorities on the task level (FTP),
• EDF assigns priorities on the job level (FJP),
• Round Robin does not assign priorities per se.

2.3. Scheduler simulator

The purpose of the scheduler simulator is to determine if a task set is schedulable by a given algorithm.
Your scheduler should always take the fastest method possible to determine schedulability and only
use simulation when it is necessary.
Simulating the execution of a task set involves computing a feasibility interval and then computing all
of the events (job releases, deadline misses, job scheduling) that occur. Referring to your lecture notes,
you have to determine what feasibility interval is applicable and use the smallest one². You can choose
whether you implement an event-based or a constant-step simulator.

¹Contact me if you want to use another language

²You can use [0, 𝑃 ) for Round Robin.

1
FACULTÉ DES SCIENCES
DÉPARTEMENT D’INFORMATIQUE

3. Running the program

3.1. Program inputs
The program should accept two mandatory arguments:
1. dm, edf or rr , the scheduling algorithm to use,
2. <task set file> , the path to the task set file.

You can add other optional arguments if you have any use for them.

$ python main.py audsley|edf|rr [-v] <taskset file> # python

$ cargo run audsley|edf|rr [-v] <taskset file> # Rust

Parsing command line arguments can be tedious. Make your life easier and use a library such as
argparse for Python or clap for Rust. It will result in a more readable code and a more robust parsing.

3.2. Process exit code

The exit code of the program will be used in automated tests to determine whether you properly detect
which method to use to determine schedulability.
The program should exit with different values depending on how you defined the schedulability of the
system. The exit codes and their meanings is shown in Table 1.

Exit code Description

0 The task set is schedulable and you had to simulate the execution.

1 The task set is schedulable and you took a shortcut.

2 The task set is not schedulable and you had to simulate the execution.

3 The task set is not schedulable and you took a shortcut.

Table 1: Exit code for each case.

Important: To specify the exit code of your program, you can use std::process::exit(<code>)
in Rust or exit(<code>) in Python. If your program does not exit with the appropriate code, it
will not pass the automatic tests and your project will be poorly graded !

4. Experiments
You are provided with a dataset of synchronous task sets with constrained deadlines described in Sec-
tion 4.1. You are required to analyse the dataset (Section 4.1) and compute the success rate of each
scheduling algorithm on this dataset (Section 4.2).
We define the success rate of a task priority assignment algorithm 𝐴 as the ratio of feasible task sets
that 𝐴 can schedule without missing a deadline. For instance, if 25% of the feasible task sets miss a
deadline, then the success rate is 0.75. This is illustrated in below diagram representing the set of all
possible task sets. The success rate is the ratio between the green area and the sum of the green and
the red areas in Figure 1.

2
FACULTÉ DES SCIENCES
DÉPARTEMENT D’INFORMATIQUE

Not schedulable by A

Infeasible

Schedulable by A
Figure 1: All possible task sets split in infeasible (grey) and feasible (others). Feasible task sets are
themselves split into schedulable and not schedulable by a specific algorithm A.

4.1. Dataset
You can find in the “Université Virtuelle” a series of synchronous task sets with constrained deadlines
grouped by utilisation and by number of tasks in the format discussed in Section 2.1. For instance, a
task set with 20% utilisation and with 6 tasks will be in the directory 20-percent/6-tasks .

For each utilisation and for each number of tasks, you have to determine the ratio of task sets that is
feasible.

4.2. Plots
• Consider the task sets in the folder 80-percent/ .
1. Plot the ratio of task sets that are feasible according to the number of tasks.
2. Plot the success rate of each algorithm (RR, DM, EDF) according to the number of tasks.
• Consider the task sets in the folders 10-tasks/ .
3. Plot the ratio of tasks that are feasible according to the utilisation.
4. Plot the success rate of each algorithm (RR, DM, EDF) according to the utilisation.

Important: An inappropriate type of plot can be anything from unreadable to misleading. Take
some time to think through which type of plot (pie, bar, line, scatter, Gantt, histogram…) is best
suited for the job. Poorly chosen plot types will be penalised.

5. Report
Write a report that includes at least (and not especially in that order)
• a short introduction that defines the scope of the project
• your methodology
• your experimental setup
• the results of your experiments alongside with the value of the parameters
• a discussion of the results
• if applicable, what you have done in addition to the project statement.

6. Evaluation criteria
The project will be graded out of 20 points balanced as follows:
• Scheduling algorithms are correctly implemented /5
• The code is readable, properly structured and uses appropriate abstractions /3
• Report /12
Total /20
You have to hand in your project as a zip file containing your code and the report on the “Université
Virtuelle” no later than the 10th of November 2024, 23:59.
Feel free to ask questions during practicals, via Teams or by email at [email protected].