About Me
I’m a PhD student at network and cloud systems research group, Max Planck Institute for Informatics (MPI-INF), working with Excellent Prof. Yiting Xia.
Before joining MPI, I got my B.S. in Comuputer Science at Beijing University of Posts and Telecommunications. I was fortunate to work with Prof. Jian Huang as a research intern at the University of Illinois Urbana-Champaign in the summer of 2020. I spent wonderful half a year as an exchange student in Institut supérieur d’électronique de Paris (ISEP) in Paris during 2019, advised by Prof. Raja Chiky and Prof. Xun Zhang.
I love building systems. My research aims to build efficient systems, especially for data centers with hardware-software codesign and cross-layer optimizations. My past research includes data center networking, systems for machine learning, and specialized hardware. I designed protocols for reconfigurable networks with data plane programmability. I also designed a specialized accelerator with FPGA for SAT solving.
I play with
- Programmable Switches (P4 + Tofino/Tofino2)
- Pytorch
- FPGA
- To be continue…
News
- I will be volunteering for SOSP’23. See you in Koblenz!
- EchelonFlow is published at HotNets’2022!
- HOHO is published at APNET’2022!
Education
- Oct 2021 - Present
PhD Student at MPI-INF - Sep 2017 - Jun 2021
B.Sc in Computer Science at Beijing University of Posts and Telecommunications, Beijing, China - Sep 2019 - Feb 2020
Exchange Student at Institut supérieur d’électronique de Paris (ISEP), Paris, France
Past Resarch Projects
Digital Molecular Computer
Digital Molecular Computer (DMC) is an in-memory computing architecture, created to solve large scale combination problems. Inspired by molecular/DNA computer, DMC combines massive parallelism of molecular computing and high speed of digital computer. Specialized microarchitecture and ISA are designed to achieve the well-organized parallelism.
We implemented the DMC prototype in FPGA and processed variable-limited boolean satisfiability problems.
For more information: Abstract Video
In-Storage Computing
The idea of in-storage computing is moving the computation to storage device to reduce data movement. The performance benefits from IO reduction but degrades from computing weakness in SSD. We modeled that tradeoff with parameters of IO, processor, dispatch model and workload metrics. The model provides a threshold of workloads’ IPB(Instruction per Byte) to determine whether a workload dispatch policy benefits from the in-storage computing system. It also provides a quantitative tool for analysis and guides us to develop dynamic workload dispatch systems in the future.
Misc.
Sports: tennis, climbing, jogging,…
In my spare time, I enjoy reading, cooking, driving, hiking and exploiting wilderness.