Have you ever had to wait for an online application to load for what seemed like an eternity? Why do we sometimes have to wait so long to make a Zoom call or to see the results of a Google search? It’s because the apps run on cloud infrastructure that is experiencing congestion.
Facebook is trying to combat such congestion by supporting cutting-edge research from renowned research institutions. This fall, Facebook announced the winners of its 2020 Faculty Research Awards. Johns Hopkins computer scientist Soudeh Ghorbani was among the recipients and will receive funding for her latest research, “Taming Data Center Microbursts at Hosts.”
“The main culprit of poor quality of service in large-scale cloud computing infrastructures and networks is short-lived congestion events called ‘microbursts.’ Unfortunately, the fleeting nature of these events—orders of magnitude shorter than the latency of the control loop of today’s network systems such as topology modifications, traffic rerouting, and rate adjustment—makes their detection and control a challenge,” says Ghorbani.
Ghorbani’s research tackles the challenge of timely detection and mitigation of microbursts by redesigning some of the foundational building blocks in large-scale networks.
“The award will support our research on identifying the major contributing factors to the formation of microbursts, revisiting some of the assumptions underlying today’s protocols that make them fundamentally slow and coarse-grained, and designing novel, high-speed algorithms and systems that are both effective in managing microbursts and provably stable and fair,” Ghorbani says.