Last updated: May 1, 2022

Insu Jang
I am a Ph.D. student in Computer Science and Engineering at the University of Michigan, advised by Prof. Mosharaf Chowdhury. I am interested in system architecture, distributed computing, heterogeneous computing (GPU, FPGA, SmartNIC, etc), non-volatile memory, and serverless computing.



  1. [SOSP ‘21] LineFS: Efficient SmartNIC Offload of a Distributed File System with Pipeline Parallelism
    [paper] [bixtex] (54/348 = 15.5%) Best Paper Award!
    Jongyul Kim, Insu Jang, Waleed Reda, Jaeseong Im, Marco Canini, Dejan Kostić, Youngjin Kwon, Simon Peter, and Emmett Witchel
  2. [ASPLOS ‘19] Heterogeneous Isolated Execution for Commodity GPUs
    [paper] [bibtex] [slides] (74/350 = 21.1%)
    Insu Jang, Adrian Tang, Taehoon Kim, Simha Sethumadhavan, and Jaehyuk Huh


Honors and Awards

  • Best Paper Award for “LineFS: Efficient SmartNIC Offload of a Distributed File System with Pipeline Parallelism”, from Symposium on Operating Systems Principles (SOSP), Oct 2021
  • Richard H. Orenstein Fellowship in Memory of Murray Orenstein, from Department of Electrical Engineering and Computer Science, the University of Michigan, 2021 - 2022
  • Korea National Scholarship, from KAIST and Korea Ministry of Science and ICT, 2016 - 2018
  • National Scholarship for Science and Engineering, from Korea Student Aid Foundation (KOSAF), 2014 - 2016

Research Projects

  • Systems for Distributed ML Training - A research project as a graduate student research assistant at the University of Michigan
    Recent trend of growing model size and parameter size forces to use multiple accelerators and distributes data across the devices. My goal is to design and implement an efficient distributed training system so that we can fully utilize all resources in data centers.

  • Reimplementing Hyperloop - A research project as a research assistant at KAIST
    Due to a copyright issue, the implementation of Hyperloop is not opened. To perform a research regarding distributed file system based on Hyperloop, we had no choice but to reimplement it by our own way.
    During implementing Hyperloop, we studied how Infiniband and RoCE RDMA works, including a key idea of Hyperloop that leverages RDMA WAIT verbs.

  • Heterogeneous Trusted Execution Environment - A research project as a graduate research assistant at KAIST
    HIX extends the protection scope of hardware-based trusted execution environment (TEE) technology to heterogenoues computing devices, e.g. GPUs or FPGAs. Based on the insight that Intel SGX protects the data in address translation (TLB entries are not inserted into the TLB for unauthorized accesses) and modern high performance device access is done through memory-mapped I/O (MMIO), we extended the protection mechanism for accessing device access through MMIO.
    Based on this SGX extension, only an authorized trusted process called the GPU enclave can access the GPU and trusted process use the GPU service through the GPU enclave via encrypted communication.