I'm interested in making networks faster, easier to manage, and more secure. I research new technologies, system platforms, and user tools to achieve these goals.
Over the last few years, my research has focused on network monitoring. My recent papers, TurboFlow (EuroSys '18, Best Student Paper) and StarFlow (ATC '18) show how to balance performance and flexibility in telemetry systems that measure traffic deep in the network core. They key is a hybrid approach, where tasks and algorithms are carefully split between hardware and software. Ultimately, the hybridization lets these systems measure terabit rate traffic at low cost while accounting for every packet and supporting custom metrics.
I'm currently working on generalizing the hybrid approach. My first step is to model the telemetry systems we've built, to better understand why they perform so well and identify other data plane functions that might also benefit from hybridization.