Understanding Power-Performance Trade-offs in Networked Applications

In this project you will measure and model power consumed by applications on a server cluster. We have developed a tool called PowerPerfCenter which can estimate expected power consumption by an application, however we need to validate our predictions with actual measurements. In this project you will carry out measurement experiments with a power-meter in a lab, and fine-tune the modeling tool to better match the measurements.

PerfCenterLite Extensions

The above tool (PerfCenter) has a complex interface which is useful only for expert users. We have developed a spread-sheet interface for the tool. This project involves enhancing that interface. It will involve understanding of performance, some python coding, and running validation experiments.

Cross-Platform Execution Time Prediction for Networked Servers

Performance analysis of websites can be done using mathematical models of the software and hardware involved. However, these models require an estimate of the execution time of a request on host CPU. Even if we measure this execution time on a test machine - how do we extrapolate this to the time on the machine on which the websit is actually going to run? This is the question you will answer. The project requires lot of experimentation - so testing and measuring on Linux machines should interest you. Check this paper to get an idea of existing work in this problem.

Autograding of Programming Assignments

We have developed a web-based application which can greatly reduce the tedium of running a typical "lab" course in programming. Check EvalPro - we are currently running a programming lab on this website. There are tons of features we need to add to this, including performance scalability, better security, crib management, and finally "program comprehension". EvalPro has also been recently merged with bodhitree so that a lecture+lab course can be run. If web-coding (using technologies such as Django Python, Javascript, AJAX, etc) is your thing, join us aboard this project and leave something lasting at IIT Bombay.