System Modeling and Simulation
2023-2024
Niveau: L2_Info Dr. Mohamed Dhouioui
Agenda
The purpose of this course is to learn the basics of systems
modeling, discrete-event simulation, and computer systems
performance evaluation.
01 Intro to modelling and simulation
02 Random numbers and Monte Carlo simulation
03 Probability models: discrete and continuous
04 Discrete-event simulation methods
05 Simulation output / input analysis
Course overview
3
Classify/describe different types of simulations.
Solve simple problems using Monte Carlo methods.
Develop discrete-event simulation models.
Generate random variates using inverse transforms.
Understand differences between transient and steady-
state behaviour of a system.
Compute confidence intervals from simulation data.
Conduct goodness of fit tests for input data models.
Use basic queueing theory for simple system models.
Basics of Modeling
4
What is a model?
— An abstract representation of a (real) system that captures the essential
characteristics or properties of the system
— Often requires making simplifying assumptions about how the system
works
Examples:
— Model airplane; molecular model; performance model
Modeling is an essential tool in computer system performance
evaluation (as we will see shortly)
Note that modeling is both an ‘art’ and a ‘science’
Modeling: A Reality Check
5
A famous quote:
“All models are wrong; some models are useful.”
- George Box, 1976
Models are useful when they provide critical insights into
the system behaviour (e.g., its performance)
Models are especially valuable when they are simple,
elegant, and computationally fast
Computer Systems Performance Evaluation
6
Performance is a key consideration in the design, procurement, and
use of computer systems.
The typical goal is to get the highest possible performance for a
given cost (e.g., dollars, energy)
Performance evaluation is a well-defined sub-domain of computer
science that has been around for 40 years
Need basic knowledge of the tools and techniques of computer
systems performance evaluation
— What are the performance requirements?
— How to compare different system alternatives?
Objectives of Performance Evaluation
7
Establish a quantitative understanding of system behavior
This understanding should be sufficient for:
Evaluating alternative system designs/configurations
— e.g., should our Web site run on one server or two servers?
— e.g., should Web server software be Apache, IIS, or nginx?
Predicting system performance for a given set of inputs
— e.g., predict the mean response time of a Web server when the number
of users is increased
Performance debugging and system tuning
— e.g., identify/remove bottlenecks, optimize configuration
— e.g., why is D2L so slow? is it the server, or the network?
Approaches to Performance Evaluation
8
Three main approaches:
1. Experimental
— Obtain measurement data by observing the events and activities on an
existing system; evaluate new algorithms or designs by implementing and
comparing them in a real system
2. Simulation modeling
— Develop a computer program that implements an abstracted model of the
physical system; manipulate the model and/or its inputs to estimate the
system performance (e.g., randomization)
3. Analytical modeling
— Represent the system by an abstract mathematical model of the physical
system (e.g., formula); manipulate parameters of the model to obtain
information about system performance
High Level Overview
9
Performance
Evaluation
Performance Performance
Measurement Modeling
Analytic Modeling Simulation
Performance Measurement
10
Measure the performance directly on a system
Need to characterize the workload placed on the
system during measurement
Generally, provides the most valid results
Nevertheless, not very flexible
— May be difficult (or even impossible) to vary some
workload parameters
Performance Modeling
11
Construct a model
— An abstracted representation of a system obtained by making
assumptions about how the system works
— Captures the most salient characteristics of the system
Reasons for using models
— Experimenting with the real system may be
too costly
too risky, or
too disruptive to system operation
— System may not even exist yet (e.g., planning stage)
Analytic Modeling
12
Mathematical methods are used to obtain solutions to
the performance measures of interest
Examples: queueing models for computer systems or
computer communication networks
Numerical results are easy to compute if a simple
analytic solution is available
Useful approach when one only needs rough estimates
of performance measures
Solutions to complex models may be difficult to obtain
Simulation Modeling
13
Develop a simulation program that implements
the model
Run the simulation program and use the data
collected to estimate the performance measures
of interest (typically using randomization)
A system can be studied at an arbitrary level of
detail
It may be costly to develop and run the
simulation program
Advantages of simulation
14
New policies and procedures can be explored without
disrupting the ongoing operation of the real system
New designs can be tested without committing resources
for their acquisition
Time can be compressed or expanded to allow for a
speed-up or slow-down of the phenomenon under study
Insight can be obtained about the interactions of
variables, and which ones have the most impact on
system performance
Can obtain answers to “What if…” questions
Areas of Application for Simulation
15
Manufacturing applications
Financial markets
Military applications
Logistics and supply chain management
Transportation modes and traffic
Business process simulation
Health care optimization
Facility placement problems
Communication networks
And many more!
Areas of Application for Simulation
16
Hair salon (e.g., Witchcraft)
Two stylists (one fast, one slow)
Limited size waiting room (N chairs)
Customers arrive at random times (no appts)
Customers are impatient (don’t like waiting long)
Question: How many customers do you lose by not having enough
chairs in the waiting room?
