CS 380N - Systems Modeling
Instructor - J.C. Browne
Modeling plays a fundamental role in the design and analysis of complex systems in general and computer systems in particular. A model is an abstraction of a system. There are frequently multiple models of a system with each model (performance model, reliability model, etc.) being specialized for analysis of a specific property. Models, like systems, are represented in languages. Algorithms and methods for analysis of models are representation specific.
The objective of the class is to enable participants to construct abstracted models of systems from which the properties of the realized system can be reliably predicted and to use these models for systems design and analysis. The principal focus will be performance models of computer systems, particularly distributed and networked computer systems. Modeling languages and analysis algorithms for other properties such as reliability, availability and correctness will also be studied. The emphasis of the course is practical although the lectures will cover theoretical concepts and principles. Participants will work with systems and tools from industrial practice and survey recent university research.
The class will be conducted on a seminar basis. The instructor will offer lectures on a set of basic topics during the first half of the semester and the participants will present reports on their projects during the second half of the semester.
3. Topics and Coverage
· Performance Models – Representations and Analysis Methods
· Queuing Models
· Finite State Machine Models
· Markov Models
· Model evaluation - Analytical, simulation and execution
· Modeling languages and systems
· Commercial tools and systems
· Research tools and systems
· Measurement and parameterization of system models.
· Model validation and validity of predictions
· Models for reliability and fault-tolerance – Performability
· Models in formal verification - Practical Model Checking
· Methods and Techniques
· Applications of Model Checking to Software
· Hierarchical modeling for physical systems
4. Course Materials
The breadth of the lecture topics precludes use of any single text. The primary sources for the course will be lecture notes, papers from the literature (both classical and recent), material from online books and tutorials and manuals for research and commercial tools. An example reference is the "Online Version of Quantitative System Performance: Computer System Analysis Using Queueing Network Models, a 1984 Text on Performance Analysis" which is available on the web at http://www.cs.washington.edu/homes/lazowska/qsp/ Several chapters from this online book are used as reference material for the lectures on queuing models and performance.
Lecture notes and papers which are not readily available elsewhere on the web will be available through the class web page.
5. Work Requirements
There will be one three hour examination over the basic topics such as model abstraction systems and model evaluation systems including analytical solution methods and discrete event simulation. This examination will be in the evening separate from class on or about the middle of March, 2002.
Each participant will also do a project. The projects may be installation and evaluation of a research language, system or tool or a case study in modeling.
· Research Systems - Each participant who chooses this option will be responsible for one research system for either measurement and/or modeling. This will involve reading the papers on the system, if possible importing and install the software implementing the system and executing a demonstration experiment with the system. Each participant will prepare and deliver one or more presentations and a written report on the chosen system.
· Case Study - The participants who choose a case study will carry though the case study using available methods and tools. There will be one or more presentations and a written report on the case study. The instructor has arranged access to projects with local industry. Possible projects include modeling of web servers, modeling of distributing computing applications, etc. Participants may propose case study projects.
Each participant will do either one or two presentations to the class on his/her project.
Grades will be derived 30% from the examination, 10% from outside assignments and 60% from the the presentations and reports on the projects.
7. Lecture Schedule and References
A schedule of lecture topics and the reference material for each lecture will be posted before the start of lectures.
Examples of lecture notes can be found on the class web page from Spring 2001.
8. Academic Honesty
The University has rules of conduct for students. It is expected that you are or will make yourself familiar with those rules of conduct and follow them.