About the course:Analyzing the performance of any computer or networked system: such as Web application servers, packet scheduling disciplines, operating system schedulers, cellular telephony networks is an important step in the design and deployment of such systems. Performance of many systems can be measured, but a sound basis in queuing system models is required for two important aspects: 1) to ensure that the performance tests are performance field data is correct, and 2) for predicting the performance of a system in a scenario that cannot be measured. The theory of queues is a mathematical theory that helps model a wide range of computing and networking systems with a common abstraction of a queuing system. In this course, you will learn the most basic results in queuing systems, in an intuitive way, and learn how to apply them to computer and network systems performance. The focus in this short course will be on being able to reason about asymptotic values of performance metrics at high loads and low loads, which is very useful in validating results of performance tests. Practical examples of performance analysis of networked servers (e.g. a Web server), of packet network links, of cellular networks etc, will be provided. A case study of interpreting the load test results of a simple Web Server using the framework of queuing systems will be covered throughout the course.PRE-REQUISITES: At least 3rd year UG in CSE Pre-requisite Courses : Operating Systems and Computer Networks.INTENDED AUDIENCE: 4th year undergraduates or post-graduates in Computer Science and Engineering, IT industry professionals engaged in computer applications performance testing and evaluation.INDUSTRY SUPPORT:Any IT company.
Overview
Syllabus
Week 1 :
1.1 Introduction, General discussion about resources and contention for resources
1.2 various types of resources, their users, and performance metrics and parameters
1.3 Intro to queuing systems - how described, standard metrics, Kendall Notation. Metrics of open queuing systems
1.4 Operational Laws, Utilization Law, Throughput, stability of a queuing system
1.5 Example using Case Study of a Load test on a web server - how the measured results match theoretical estimates, where do they not match.
Week 2 :
2.1 Asymptotic Analysis of G/G/1, G/G/1/K queues: Values of metrics at low load and high load asymptotes
2.2 Asymptotic Analysis of G/G/c/K queues, Examples
2.3 Little's Law - Intro and discussion
2.4 Little's Law - further discussion and Visual "Proof"
2.5 Little’s Law examples: Continuing the Case Study of a Load test on a web server - - how the measured results match theoretical estimates, where do they not match.
Week 3 :
1.1 Closed Queuing Systems. Metrics, parameters. Analysis of simplest closed queueing system
1.2 Closed Queuing System: Low Load and High Load Asymptotes of all metrics
1.3 Response Time linear asymptote, Kleinrock's Saturation Number Heuristic
1.4 Closed Tandem Queue Low Load and High Load Asymptotes of all metrics, Response Time linear asymptote, Kleinrock's Saturation Number Heuristic
1.5 Closed queuing system examples: continuing Case Study of a Load test on a web server - - how the measured results match theoretical estimates, where do they not match.
Week 4 :
4.1 Closed queuing network with branching and feedback
4.2 Closed queuing network analysis (example): visit count, all metrics, asymptotes
4.3 General formulation of Jacksonian Closed Queuing Networks
4.4 Arrival Theorem, Mean Value Analysis (Derivation)
4.5 Mean Value Analysis examples: concluding Case Study of a Load test on a web server. Discuss applications and limitations of queueing systems based modeling
Taught by
Prof. Varsha Apte
