Vasily Tarasov wins IBM Ph.D. Fellowship for 2012

Go Back to News Highlights

News Headlines

Computer Science PhD student Vasily Tarasov had won an IBM PhD Fellowship for 2012, second year in a row.



DETAILS

IBM Ph.D. Fellowship recipients are selected on the basis of outstanding research and technical excellence in areas of interest to IBM. These competitive fellowships provide a stipend, tuition, and fees to the awardees. The fellowships are also structured to allow students to pursue a technical career in either IBM's Research Division or IBM's development laboratories.

This is a highly prestigious annual competition. Less than 10% of world-wide applicants make it to the 2nd selection phase, and fewer make it to the final award. Often, schools are also limited to submit no more than 1-2 names annually.

Professor Zadok's student, Vasily Tarasov, studies the impacts of the virtualized workloads on the modern multi-layer storage stack. One of the major trends in modern Information Technology is a shift from traditional physical hosts to virtual environments that share a powerful server. I/O workloads created by virtual environments are significantly different than ones created by physical machines. In addition, as I/O requests pass through multiple layers in a virtualized storage stack, their characteristics change drastically. As a result, a workload observed by the bottom-most I/O layer is quite different from the original workload observed at the upper-most application layer.

Tarasov's work proposes to collect traces from multiple layers in a virtualized storage stack, analyze them using a custom developed Trace2Model tool, and characterize the changes observed. Such a characterization is a first step towards designing new storage systems that are optimized for virtualized workloads. The final goal is to provide better performance of scalable Network-Attached-Storage (NAS) solutions, where the back-end file system is a distributed file system, such as IBM's GPFS. Only such solutions will be able to keep up with the scale of the future VM deployments as seen by modern Clouds.


Department of Computer Science • Stony Brook University, Stony Brook, NY 11794-4400 • 631-632-8470 or 631-632-8471