Research

My research interests are in the area of computer systems modeling and performance evaluation. My earlier work largely focused on (a) development of models and efficient methodologies for evaluation of systems at design time - specifically, on techniques that lead to reasonably simple abstractions (needed for ease of applicability) and efficient techniques (needed for rapid evaluation of design ideas), and (b) quality-of-service-based design of large-scale continuous media (CM) storage systems subsequent performance evaluation - efficient storage system designs are fundamental to the viability of a broad range of systems and applications.

More recently, one aspect of my research focused on design and evaluation of large-scale peer-to-peer (p2p) and overlay systems. Addressing challenges in p2p and overlay systems can lead not only to significant improvements, but also to significant insight - not only into p2p and overlay systems but into distributed systems in general. Such understanding and subsequent ability to provide quality-of-service (QoS) and performance prediction relies on thorough evaluation. Doing such evaluation through analytical modeling can lead to (a) quick evaluation of design-time choices and (b) insight through better understanding of systems. Hence, my work in this area has largely focused on (i) development of accurate analytical models and (ii) design of approaches that can lead to QoS predictions.

Another aspect of my research has been reliability evaluation of software systems. High level abstractions that allow for estimating reliability of software architectures (rather than implementations) can provide insight into the reliability characteristics of a software's design while facilitating timely and less costly design improvements. % Given the complexity of modern software, scalability of the evaluation methodology is a critical concern. Moreover, wide-spread adaptation of such techniques by software engineers (largely) hinges on them ``fitting into'' existing design processes. % Hence, I work on analytical reliability evaluation methodologies that (a) are easily derivable from architectural designs and (b) lead to efficient solution techniques, needed for rapid evaluation of design ideas.

I have also focused on analysis of data generated by wireless sensor systems. Such systems can facilitate scientific studies by instrumenting the real world and collecting corresponding measurements, with the aim of detecting and tracking phenomena of interest. Given the huge volume of measurements such deployments produce, accurate and robust automated processing of data is fundamental to its meaningful use. Thus, I expect our work to be of use to a broad range of current and future sensor systems and their applications.

More recently, I am exploring problems related to the scale and complexity of highly distributed data intensive systems, where one direction is exploration of synergy between machine learning and system modeling.