Varsha Apte

Professor, CSE, IIT Bombay

Have you run a load test and generated a ‘response time vs number of users’ plot like this? How many metrics and parameters of the system can be predicted by just observing this graph? Turns out, you can deduce a lot if you know how to apply queueing theory to your load testing results. Knowing queuing theory can help make sense of load testing results, quickly spot experimental errors, and shorten the time to complete tests due to clever planning of tests. In this training event co-located with ICPE 2019, in the morning session, you will be introduced to the fundamental results from queuing theory, and will learn how to apply them during load testing. Numerous exercises will be given to reinforce your learning. In the afternoon session, you will be given a crash tutorial on programming of discrete event simulations for performance analysis, after which you can take part in a Simulation Coding Hackathon.

Performance Sherlock (Morning Session)

Outline

1. Introduction to performance metrics and parameters
2. Introduction to Open Queuing systems
3. Asymptotic Analysis of Open queuing systems
4. Basic Laws of queuing systems (Little’s Law, etc)
5. Introduction to Closed Queuing systems
6. Asymptotic Analysis of Closed queuing systems
7. Mean Value Analysis
8. Practice exercises

Who can benefit from this session

Anybody who has not formally learned queuing theory, but is engaged in load testing or other performance analysis work in his/her job or research project. It is targeted towards ‘beginners’ - either professionals starting out in the industry, or students carrying out systems research work. Participation certificates will be issued to all who attend and participate in this session.

Who can participate

ICPE 2019 attendees who register for either the 3-day ICPE conference, or for the full conference+tutorial/workshops are invited to register free for this session.

Simulation Modeling Hackathon (Afternoon Session)

This is a follow up to the morning session. In this session, the method of discrete event simulation for performance analysis will be introduced briefly, and a sample program written in C for simulation of a single server queuing system will be explained and demo-ed. After that, attendees may work, preferably in teams of two, to incorporate some advanced features in the simulation program, and carry out some interesting performance analysis using this program. They will be given the description of a system and asked to write a new program or modify the given program to simulate the system. They should pose some interesting design or analysis questions which will be answered by the simulation program. A very brief report, based on the analysis done will be submitted by each team. The team which submits the best report before Monday April 8th, 4pm, will be given a certificate of special mention (at the end of the conference) . Details regarding how the submissions will be judged will be posted here soon.

Who can benefit from this session

Familiarity with the material from Session 1 will be assumed for this session. Anybody who is engaged in a somewhat deeper performance analysis study of a product, protocol, etc will benefit. Participation certificates will be issued to all who attend and participate in this session.

Who can participate

ICPE 2019 attendees who register for either the 3-day ICPE conference, or for the full conference+tutorial/workshops are invited to register free for this session. It is preferable to bring a laptop to the Hackathon session.

Sessions designed and Led by Varsha Apte

Registration is required for this event by filling this form . Each session will be held only if at least 15 participants register, and a maximum limit will also be enforced. Registration Deadline - March 15th, 2019. Confirmation will be given by March 18th, 2019.