INNOVATIVE METHODS AND TECHNIQUES
FOR HIGH-PERFORMANCE AND - RELIABILITY
MODELING AND SIMULATION (M&S)
Prof. Dr. Axel Lehmann
Institut für Technik Intelligenter Systeme (ITIS)
Institut für Technische Informatik
at the
Universität der Bundeswehr München
Universität der Bundeswehr München
Germany
Germany
e-mail: [email protected]
URL: http://www.informatik.unibw-muenchen.de/inst4/
OUTLINE
1. Modeling & Simulation (M&S):
“State of the Art” and Demands
2. M&S: A Multiple-Phase Development & Application Process
3. Component-Based M&S
4. M&S - Verification, Validation & Accreditation (VV&A)
5. Parallel and Distributed M&S
1. M&S - “State of the Art” and Demands
Trends / Requirements of (Technical) Systems Developments
•
Rapid technological innovations
Æ
new information and communication technologies
Æ
efficient,
powerful,
computer-assisted tools (e.g. CAD, CAM, . . . )
•
Increasing systems complexity & lifetime
Æ
embedded
systems
Æ
distributed
systems
Æ
networks of components / systems
•
Increasing productivity & cost-benefit
Major
Challenges
•
“Mastering” of system(s) complexity
over lifetime w.r.t. multiple aspects /goals?
(safety, reliability, performance, ...)
•
“Mastering” of model(s) complexity
!?
⇒
STRATEGIC approach: “Devide and Conquer” !!
Current Importance of M&S
(as a discipline / methodological approach / tool set):
Æ
Receives increasing acceptance by decision makers
Æ
Becomes more and more a “standard” method / tool set
Æ
Seen as a major enabling technology for innovations
• 1995, US-DARPA:
“... M&S is one of the top-10-key enabling technologies ...
• 1998, DoD (Dr. Gansler): “... by the year 2000 ... Systems development in 25 % less time...”
• 1999, US-government:
IT
2Research Initiative
• 1999, PITAC report:
“Fund research in ... global-scale networks and its associated
information infrastructure .... including .... Modeling and
simulating network behaviour (Recommendation 3.3.2)
Major Challenges for M&S applications:
• Increasing systems / M&S complexity
• Decreasing cycle times for systems / M&S innovations
• Increasing lifetimes of systems, models, and simulations
• Increasing variety of M&S-aspects / purposes
• Safety, reliability, ... cost-benefit constraints (for systems / M&S)
• “Hardware-/Software-/User-in-the loop” simulation
• User acceptance; ease of use & credibility
General Approaches, Methods to cope these
M&S-Challenges:
• Hierarchical modeling
• Hybrid models (OR-, analytic, simulation solutions)
• Interoperability of models
• Reuseability of models (model components)
• Collaborative M&S
• Parallel & distributed M&S
• Improved model credibility by application of VV&A
• Model engineering
2. M&S - Design, Implementation and Application
Process
Example: Effectiveness and efficiency of a “Booking System”
(Customers who access their bank accounts to transfer or to deposit
money; client-server architecture)
Goal parameters to be analysed might be, e.g.:
Æ
processing time per transaction
Æ
client/server utilization
Æ
queueing time, queue length
How to approach this (complex) problem?
⇒
“Divide and Conquer”:
Phases & Products in the M&S
-Development Process
Model Input Data
Solution Techniques
Model Documen-tation Objekt Objekt Objekt Objekt Intera ktion Systemgrenze “Umwelt-objekt“ “Hauptobjekt” mit ModellattributenCommunicative
Conceptual Model
, Project ObjectivesStructured
Problem Description
Formal
Model
Executable
Model
Model
Results
Problem
Definition
System
Analysis
Model
Formalization
Implementation
Experimentation
Modeling Method
System Observations
Examination Aim
Experimental req.&constr.
Technical req.&constr.
Formal req.&constr.
Conceptual req.&constr.
M&S-Sources of Knowledge and
Expertise
Problem
Definition
System
Analysis
Model Formalization
Implementation
Experimentation
Modeling
Expertise
(HW-)SW-Expertise
Experimental
Design and
Domain
Knowledge
User
Knowledge
Problem
Definition
System
Analysis
Model
Formalization
Implementation
Experimentation
Project Manager
(Contractor)
Modeller
Domain Expert
User
Modeling
Expertise
Experimental
Design and
Analysis
Domain
Knowledge
(HW-)SW-Expertise
User
Knowledge
Customer
Æ
“Divide and Conquer”-approach:
decomposition of (complex) models/problems
→
“components”
Æ
“Component” : not a well-defined term !!
Example: “Booking System”
Example: „Booking System“
Conceptual Model :
I/O device 1
CPU
T 1
T n
I/O device n
• • • • • •Server
Example: Booking System
Example: Booking System
Formal Model
(Version 2):
Little‘s law:
w
Q
t
k
=
λ
or
=
λ
Performance measures:
With: response time t, queueing time w, service rate µ
t = w + ; k =
µ
1
Σ
k • p (k)
6
State probability
p(k)
6
Utilization
ρ
( m service stations)
1
<
=
=
µ
λ
ρ
m
service rate
arrival rate
Conclusions
↔
“model component”
(e.g. regarding the example “Booking System”)
•
Problem Description:
Æ
pragmatism / goal specification
•
Conceptual Model:
Æ
structural & functional description of
“components”
Æ
different levels of abstraction
•
Formal Model:
Æ
formal specification of
“components”
(
↔
selected modeling paradigm(s))
Æ
different levels of abstraction
Æ
hierarchical modeling approach
(
•
decomposition into submodels /
“components”
•
Executable Model(s):
e.g.
Æ
analytic solution reusable SW-
“components”
Phases & Products in the M&S
-Development Process
Model Input Data
Solution Techniques
Model Documen-tation Objekt Objekt Objekt Objekt Interaktion Systemgrenze “Umwelt-objekt“ “Hauptobjekt” mit ModellattributenCommunicative
Conceptual Model
, Project ObjectivesStructured
Problem Description
Formal
Model
Executable
Model
Model
Results
Problem
Definition
System
Analysis
Model
Formalization
Implementation
Experimentation
Modeling Method
System Observations
Examination Aim
Experimental req.&constr.
Technical req.&constr.
Formal req.&constr.
Conceptual req.&constr.
X
X
X
X
X
6
Model Federation Level
„Black Boxes“
6
Model Level
Autonomous, interoperable
models
6
Submodel / Object Level
Submodels /
Object structures of
different modeling
paradigms
6
Function Level
Coded basic functions /
algorithms
( ) 6 0 und 0 mit : 0 0 ) ( wenn : ) , ( ) ( ) ( . 1 1 1 1 ≤ < ≤ < − ⋅ ⋅ − + > = ∀ + + + + s n i sonst t l t t z t t r f t l t l s s i si s i i si i si i s
Model („Component“) Specification
Levels
Library of
submodels &
communication
infrastructure
Library of
objects/methods (for
interaction)
Program Library
Model
repository
Component-Based M&S : Current Approaches
•
Hierarchical modeling via
- decomposition
- aggregation
- hybrid solution / implementation techniques
•
Generic modeling object templates
(depending on the modeling paradigm), e.g.
- class / object libraries
•
Function / Program libraries, e.g.
- statistical analyses
- random number generators
•
Coupling of monolithic models, e.g.
- federation of models (DIS, HLA, . . . )
- agent-based simulation
Conclusion: Missing
comprehensive
formal &
4. M&S-Verification, Validation &
Accreditation (VV&A)
Terminology
6
Model
validation:
process of demonstrating that a model and its behavior a
suitable representations
of the real system
and its behavior
w.r.t.
intended purpose
of model application.
6
Model
verification:
process of demonstrating that
a model is
correctly represented
and was
transformed correctly
from one representation
form into another,
w.r.t.
transformation and representation rules
,
requirements, and constraints.