Web applications, resource sharing and performance prediction
Imperial College London
DEI - Seminar room
3rd May, 2011
The on-demand enterprise computing model is expected to become mainstream in the next years. Web-based enterprise software will be mainly offered as a service to multiple client organisations from large hosting platforms, such as private or public clouds.
This new computing model raises the need for service providers to deliver predictable performance, understand software and hardware bottlenecks, and stress-test the resilience of applications to time-varying workloads and their peaks. However, this is a difficult problem, since the performance of Web-based applications is determined by the many interactions between the incoming requests and the software components and shared resources that serve them.
In this talk, I will discuss a performance engineering framework for Web applications. Based on live system measurements, the proposed methodology automatically derives a special class of tractable hidden Markov models to characterize resource consumption at different levels. Queueing-theoretic extensions are then defined, solved by
numerical optimization programs, and used to predict, characterize, and test Web application performance.
Giuliano Casale received the MSc and PhD degrees in computer engineering from Politecnico di Milano, Italy, in 2002 and 2006 respectively. He joined the Department of Computing at Imperial College London in 2010 where he is currently an Imperial College Junior Research Fellow. His research interests include performance modeling, workload characterization, stochastic scheduling, simulation, and resource consumption estimation. Prior to joining Imperial College London, he was a full-time researcher at SAP Research UK in 2009, and a postdoctoral research associate at the College of William and Mary, Virginia, between 2007 and 2008. In Fall 2004 he was a visiting scholar at UCLA studying bounds for queueing networks. He has published more than 60 papers in journals, conferences, and book chapters. He is a member of the ACM, the IEEE, and the IEEE Computer Society.