BenchLab: Benchmarking with Real Web Applications and Web Browsers
Event details
| Date | 06.04.2011 |
| Hour | 14:30 |
| Speaker | Dr. Emmanuel Cecchet, University of Massachusetts |
| Location | |
| Category | Conferences - Seminars |
Popular benchmarks such as TPC-W and RUBiS that are commonly used for evaluation by the systems community are no longer representative of modern Web applications. Many of these benchmarks lack the features such as JavaScript and AJAX that are essential to real Web 2.0 applications. Further, traditional benchmarks rely on browser emulators that mimic the basic network functionality of real web browsers but cannot emulate their more complex interactions. Rather than proposing a new benchmark with a web application and browser emulators that try to approximate real applications, we propose to use real browsers with real applications and datasets. We have rebuilt the Wikipedia software stack with multiple real datasets (Wikibooks, Wikipedia in different languages) and collected real traces from the Wikimedia foundation. In this talk, we will present BenchLab, an open source framework that allows capturing live web site traces and replaying these traces using real web browsers (Firefox, IE, Chrome) deployed anywhere on the Internet. BenchLab can also be used to distribute benchmark applications, traces, experimental results and reproduce entire experiments. BenchLab provides already packaged virtual machines containing applications, databases and web browsers for researchers to experiment with Internet scale benchmarking of real applications using private or public clouds. We will present early results with Wikibooks and perform a demo of the software if technically possible.
Links
Practical information
- General public
- Free
Contact
- Madeleine Robert