Sushovan Das
I am a postdoctoral researcher at ETH Zurich in the group of Prof. Laurent Vanbever. Prior to ETH, I got my PhD from Rice University, Dept. of CS, advised by Prof. Eugene Ng. Previously, I pursued Master’s at IIT KGP, India, in the Dept. of E&ECE, under the guidance of Prof. Debasish Datta.
My research primarily focuses on designing novel and energy-efficient optical network architectures to optimize diverse application performance in future-generation cloud infrastructures. Today’s cloud infrastructures have become ubiquitous across the globe hosting diverse applications e.g., machine learning, deep neural network, database, etc. Such large-scale applications are distributed across hundreds of servers, requiring inter-server communication (data exchange) several times during the execution. Recent advancements in domain-specific accelerators such as GPU, TPU, non-volatile memory etc., shrink the computation time significantly; hence communication has become the major bottleneck. Although, current silicon-based commodity switches would consume high energy and cost with increasing speed. Under such energy-critical situations, designing high-performance next-generation cloud leveraging low-power and low-cost optical switches, seems to be the most promising alternative. My vision is to design holistic all-optical network architectures for the cloud [Shufflecast, OSSV, RDC] that can efficiently handle diverse traffic patterns at scale.
During my Master’s I have worked on theoretical performance analysis (both physical and MAC layers) of optical WDM-packet switched metropolitan area networks. Additionally, I have pursued an internship at Accenture Labs under the guidance of Dr. Sanjoy Paul, where I worked on designing low-latency control and data planes for next-generation cloud-native 5G core networks. This was a collaborative work with WINLAB, Rutgers University.
Going forward, I am primarily interested in exploring unique use cases for optics in diverse networking applications/primitives of cloud/wide area/satellite networks and designing novel algorithms/systems.