System Concepts

1. A system is an integrated whole (as opposed to a

collection of parts) manifesting a behavior or a
functionality.
2. A system is a set of elements dynamically interacting
and organized in relation to a goal" (J. de ROSNAY,
1990, p. 93).
3. "System" corresponds to a man-created general
conceptual model for coherent, complex and more or
less identifiable and permanently observed real world
entities.
4. There are various degrees of systemic organization,
from the least integrated (composite or distributed
systems) to the most differentiated and integrated
ones.
5. Consider de Rosnays definition. Must a system have
a goal? A finality? What is the goal of a man? What
is the goal of an ecosystem?
6. If a system is a contained collection (meaning its
bounded by an environment), how does it expand?
Note that a closed structure cannot grow, and eventually
would collapse internally.

7. A system is a set of parts with a common destiny,
which maintain their interrelations, even when placed
in a different environment" (F.BONSACK, 1990)
What is a common destiny? It means a functional unity in
a systemic sense if all the parts work as they are supposed
to do, the whole system works as well.
8. A system is a set of interrelated elements.
This implies there must be at least two elements in the
system and they must be in interaction. Also, no subset of
System
Thread I
Interface
provider
elements is unrelated to any other subset (no
disjointedness).

9. Systems are most generally characterized by their
complexity, their coherence and relative permanence,
and their tendency to seek their own survival. These
general conditions dominate the whole concept.
System characteristics
1. Every system should do or deliver what it is supposed
to do. (Performance issue)
2. One customer at a time? Multiple customers at a time?
Different architectures. (Service/goal issue, survivality
issue)
Disjointed
systems
Thread I
Interface
provider
Multithreaded
3. Single or redundant server? Units with multiple
redundancy?
How do we define redundancy? How about this?
Redundancy is the probability that the system is working
(can deliver what it is supposed to do) at a time t t +
given that it was working at time t.
System modeling = Modeling = Problem Solving
System design = Specification of a model in terms of its
components (architecture), expected behavior profile
(performance)
Server
Queue
Service requesting
Clients
One customer at a
time organization
Single thread
Thread I
Interface
provider
Thread J
Computation
Thread K
Pipeline thread
Thread L
Communication
Thread M
Cache manager
Thread N
Controller
Thread P
Buffer manager
Thread Q
Miscellaneous
Buffer
Server

Any better solution?
Static System? Dynamic System?
Sequential/Concurrent?
Determinate system? Non-determinate? Adaptive? Error-
correcting?
Classical System? Quantum System?
4. System should have stable operating points (stable
equilibria) but at the same time it must be adaptive to
new changes or challenges. A difficult problem to
tackle. Why?
For any dynamical system of the form
) t , y ( f
dt
dy

,
The equilibrium points are at
0 ) t , y ( f
dt
dy

System
Extn
trigger
response
Null-redundant
System
Primary Server
Secondary Server
Secondary Server
Secondary Server
response
Extn
trigger
Multi-redundant
System
EMBED Equation.3
Multiple equilibrium points. Some of them are stable
equilibriums, some unstable equilibriums, some limit
cycles.

Systems are often studied and approached through
simulation. Simulation is not a must, but helps as an aid .
0 y
time
y
unstable
stable
What is Simulation?
A Simulation is the imitation of the operation of
a real-world process or system over time.
Could be done by hand or on a computer.
Why simulation?
Granularity issue. How good is the estimated
system?
Involves generation of data & artificial
history of a system, observation of the data and
history, and inferences concerning the systems
profile.
How high is the level of confidence?

0
H
Proposed model (
?) version

Shall we accept or reject this at some level

? We can have two types of errors:

Actual
Performance
Simulated
systems
Performance
Time
A type I error is rejecting the proposed
model
A type II error is accepting the model
when the model is a false one from its manifest
behavior or from its extensive sampling
To study a system, we often have to make
assumptions about the operation of the system.
May not always be correct. Future versions might
do a better job, etc.
Assumptions about its performance and the
attendant architecture constitute a model, used to
understand the behavior of a system.
Simulation vs Analytical solution?
Analytical solutions: if the model relationships
are simple enough to use mathematical methods
to obtain exact information on system
Simulation: Develop a Simulation model and
evaluate the model usually with a computer to
estimate the desired characteristics of the
model.
Simulation Model
A simplified representation of a sys. (or process
or theory), not sys. itself.
Models cant have all attributes; they are
simplified, controlled, generalized, or idealized.
For a model to be useful, all its relevant
behaviors must be determined in a practical way,
given a reasonably limited set of descriptions.
A model must be validated.
After validation, a model can be used to
investigate and predicate system behaviors, or
answer what-if questions to enhance
understanding, training, prediction, and
evaluation of alternatives.
When Simulation is Appropriate?
Study the internal interactions of a complex (sub)-
system.
Observe models behavior and resulting outputs due
to changes on
external environment or internal variables.
Improve system through model building.
Experiment new designs and policies prior to
implementation.
Understand & verify analytic solutions.
Identify & determine requirements.
Allow training & learning at a lower cost.
Visualize operations through animation.
Its difficult, time-consuming, expensive, hazardous,
or impossible to solve the problem by conventional
analytic or numeric methods.
When Simulation is Not Appropriate?
When the problem can be solved using common
sense.
When the problem can be solved analytically.
When it is easier to perform direct experiments.
When the costs exceed the savings.
When resources or time are not available.
When no data is available.
When verification & validation cannot be performed.
When the power is overestimated.
When the system is too complex or cant be defined.
Areas of Applications
Computer systems, e.g. scheduling, memory
management
Computer networks and communications
Manufacturing
Semiconductor manufacturing
Construction engineering
Military applications
Transportation
Business process
Systems and system environment
To model a system, we first need to know what
a system is. How far does the system extend?
A system: objects and their relationships and
interactions. How do we simplify or model
interactions?
System environment: changes happen outside
the system, but affect the system

Boundary (Porous? Dense? Stable? Unstable?)
Components of a System
Entity: system object
System
Abstraction
Attribute: property of an entity
Activity: operation is a specific time period
State: collection of variables needed to describe
the system at any time.
e.g. System model S. Attribute space
X x
. Activity
space
A a
. State space
S s
. System profile over a
time-interval: initial state
0
s
then
p 2 1 0
s s s s ...
Event: an instantaneous occurrence that may
change the state of the system.
e.g. The system changes the state from 1 p
s
to p
s
if at a
state 1 p
s
the event is
E e
. Not every event causes a
system change. We assume an action or an event if
system changes its state.
Discrete and Continuous Systems
0
a
1
a
2
a
3
a
4
a
E e a
i i

Discrete system: state variable(s) change only at
a discrete set of points in time.
Example: number of customers waiting in line

+1 n
X
Number of customers the departing customer
at time n leaves in the system
=

'

+
0 X if A
1 X if A 1 X
n
n n

where A = arrivals during the service time of the
(n+1) the customer
Continuous system: state variable(s) change
continuously or smoothly over time.
Example: chemical level in a tank, electric
current
Server
Solution of a viscous fluid-flow may obey
(Hydrodynamic equation primarily known as
Navier-Stokes equation)

0 u
t
+

) .(

, ) . ( u p
1
u u
t
u
2
+ +

0 u p T K u
t
H
+

,
_

. ) .( .

A continuous model.
Types of Models
Time
Voltage across
capacitor plates
V
Models:
Physical: model home, model of a bridge, wax
model of a person

Mathematical (symbolic):
In Mathematica: Symbolic differentiation yields

In[1] =
] ], log[ ^ [ x a x 2 x D +
Out[1]=
) ln( a x x 2
x a
x
2
+ +
+

Simulation model
Static (at some point in time) vs. Dynamic (change
over time)
Deterministic (known inputs) vs. Stochastic
(random variables, inputs/outputs)
Also, another classification.
Analog vs. Digital
Analog
Good
o High bandwidth
o High resolution
o Specific control functions are available as off-
the-shelf ICs
o Analysis and design methods are well-known
Bad
o Temperature drift
o Component aging
o Sensitive to noise
o Hardware design
o Can implement simple designs only
o No communication capability
Digital
Good
o Programmable solution
o Less sensitive to environment
o Can implement advanced control algorithms
o Capable of self-tuning, adaptive control, and
nonlinear control functions
o Communication capability
Bad
o Data converter is required.
o Analysis and design methods are more complex
o Sampling & quantization error
o Computation delay limits the system bandwidth
Focus of this course:
Discrete vs. Continuous
Discrete, dynamic, and stochastic
Discrete-Event System Simulation
Modeling of systems in which the state variable
changes at a discrete set of points in time.
Methods: numerical instead of analytical
Analytical: deductive reasoning of math; accurate
Numerical: computational procedures; approximate
Simulation models are run rather than solved.
Observation of the real system, entities, interactions
Assumptions of the model
Collection of data
Analysis and estimation of system performance
Steps in a Simulation Study
Problem formulation
Setting of objectives and project plan
Model conceptualization
Data collection
Model translation
Verification
Validation
Experimental design
Production runs and analysis
More runs
Documentation and reporting
Implementation
Verification and Validation
The most important step in the process: validation
Validation should not be an isolated task that follows
model development, but rather an integral part of
model development.
Verification: Are we building the model right?
Is the model programmed correctly (input
parameters and logical structure)?
Validation: Are we building the right model?
Is the model an accurate representation of the real
system world process or system over time?
Iterative process of comparing the model to actual
system behavior and refine the model.

Model Building
Iterative process consisting of three main steps:
Observe the real system & the interactions of
components and collect data
Domain specific knowledge
Stakeholders: operators, technicians, engineers,
Construct a conceptual model
Assumptions or hypotheses on components and
parameter
values
Structure of the system
Translate the operational model to computer
recognizable form
An example. Simulation to observe movement of
radioactive particles in a densely packed medium. A lead
blanket of thickness d.
Model.
Upon hitting a nucleus, it can be
a. completely absorbed with a probability p
b. scattered with a probability (1-p) to a new site
If (b), move to the new site at a random direction , and a
random throw-length l in the direction of .

Net movement in the x-direction is
cos l x x
old new
+
In terms of this, picture the problem as
nucleus

l
x-direction
Jacket
Surely, whether a particle would escape the confine of a
jacket would be predicated by p, the capture probability.
What kind of behavior do we expect? What would be the
minimum jacket thickness to capture 95% or more of the
radiation?
Typical questions:
Should we treat incoming particles as independent
particles? Should how many be coming within next t
units of time depend on how many already entered?
Should we spice it up by assuming a max number of
collisions a particle can withstand?
References:
Th. Baeck: Evolutionary Algorithms in Theory and Practice: Genetic Algorithms,
Evolution Strategies, Evolutionary Programming, Oxford University Press, New York
1996.
D. Ballard: An Introduction to Natural Computation. The MIT Press, Cambridge, MA,
1998.
J. Banks (Ed.): Handbook of Simulation. John Wiley & Sons, New York, 1998.
M. Berthold, D.J. Hand (Eds.): Intelligent Data Analysis. Springer, Berlin, 1999.
H. Bossel: Modeling and Simulation. Vieweg, Wiesbaden, 1994.
V. Cherkassky, F. Mulier: Learning from Data. Wiley, New York, 1998.
N.A. Kheir (Ed.): Systems Modeling and Computer Simulation. Marcel Dekker, New
York, 1988.
A.M. Law, W.D. Kelton: Simulation Modeling & Analysis. McGraw Hill, New York,
1991.
H.-O. Peitgen, H. Juergens, D. Saupe: Chaos and Fractals. Springer, New York 1992.
S.S. Rao: Engineering Optimization: Theory and Practice. Wiley, New York 1996.
M. Schroeder: Fractals, Chaos, Power Laws. W.H. Freeman and Company, New York
1991.
H.-P. Schwefel: Evolution and Optimum Seeking. Wiley, New York, 1995.
A. G. Wilson: Catastrophe Theory and Bifurcation. Groom Helm Ltd., London, 1981.