My current research deals with improving the performance of large-scale distributed systems. More specifically, I developed a custom large-scale distributed system that I am using as a research tool to study the performance implications of system-level bottlenecks as well as software design choices that impact such distributed systems.
Network latencies have become increasingly important for the performance of web servers and cloud computing platforms. My work deconstructed the “tail at scale” effect across TCP-IP, UDP-IP, and RDMA network protocols.
False sharing causes a serious performance loss in multi-threaded applications that run on multicore systems. My research deals with detecting cache contention caused by false/true sharing with a very low overhead, and incorporating an online mechanism to mitigate performance degradation caused by such sharing patterns.
I developed a tool for online detection and mitigation of performance degradation caused by the four kinds of cache misses. This work was the goal of my internship at Intel Labs, where Dr. Gilles Pokam mentored me.