Research Statement Sang-min Park

Across industry and academia, people are witnessing the accelerating transition from small-scale, closed, inhouse computing and data storage to mega-scale, open, and service-oriented infrastructure, called Cloud computing. In the last few years, as a graduate student at the University of Virginia, I am conducting research that can make a significant impact to this emerging field. Specifically, I have been addressing resource scheduling problems to deliver more predictable computation to applications while resource utilization on the shared infrastructure is sustained at a high level, thereby satisfying both end users and infrastructure operators. The outcome of my research has been promising; my thesis shows that the application of rigorous scheduling theory (i.e., control theory) to virtualized resource containing mechanism (i.e., VM scheduler) can deliver predictable computation to important applications, including eScience and MapReduce programs. Since 2002, my personal pursuit has been in the area of high performance, distributed systems. My approach to such research problems is to design a potential solution and then to prototype a working system, albeit sometimes an academic, small-scale one. While pursuing a PhD, my passion in research and software development has incorporated sound principles and rigorous tools. Finding an important research problem, understanding the problem to the deepest extent, and personalizing the problem by emotionally committing to it have become my guiding principles. The application of sound tools (e.g., control theory), which is not usually used in my area, has been highly effective in addressing the issues that are increasingly important regarding Cloud computing. Below, I address why scheduling problems are important in large-scale, shared infrastructures; present my control-theoretic, virtualized approach in detail; and sketch out my future research directions.