ORIE Colloquium: Daniela Hurtado-Lange (Northwestern)

Performance Analysis of Data Center Networks: Drift method and Transform Techniques

Today’s era of cloud computing and big data is powered by massive data centers. The focus of my research is on resource allocation problems that arise in the design and operation of these large-scale data centers. Analyzing these systems exactly is typically intractable, and a usual approach is to study them in various asymptotic regimes, with heavy traffic being a popular one. In this talk, I will illustrate two methods for heavy-traffic analysis that we developed in my research. First, I will present the drift method to study scheduling in data center networks. I will present this result in the context of the so-called generalized switch, which subsumes several resource allocation problems in data centers. Second, I will present the use of novel transform techniques to characterize the tail behavior of delay for efficient load balancing of jobs on servers. The use of these techniques in non-heavy-traffic regimes will also be presented.

Bio:
Daniela Hurtado-Lange is an assistant professor in operations in the Kellogg School of Management at Northwestern University. Before joining Kellogg, she spent nearly two years as an assistant professor of mathematics at William & Mary. She got her Ph.D. in operations research at Georgia Tech, under the supervision of Prof. Siva Theja Maguluri. She got her undergraduate degree in industrial engineering and mathematics at Pontificia Universidad Católica de Chile, followed by an M.S. in industrial engineering under the supervision of Professor Pedro Gazmuri. Her research was recognized with the second prize on the INFORMS 2020 JFIG paper competition, and the 2022 Sigma Xi Best Thesis Award in Georgia Tech. Her research interests are broadly on applied probability, specifically on performance analysis of stochastic processing networks.