Bio#

I am a third-year Ph.D. student in the Operating Systems Laboratory at Tsinghua University, advised by Prof. Yu Chen. My research focuses on operating systems, including scheduling, networking, and storage. I study hardware-software co-design and kernel-bypass techniques to improve system performance.

Education#

  • Tsinghua University, Ph.D. in Computer Science and Technology, 2023.8 - present

    Recipient of the 2025 Comprehensive Scholarship. Expected graduation: June 2028.

  • Tsinghua University, B.Eng. in Computer Science and Technology, 2019.8 - 2023.7

    Recipient of the 2019 Freshman Scholarship, 2020 Academic Excellence Scholarship, and 2022 Comprehensive Scholarship.

Research#

  • Kaifu Tian, Youjie Zheng, Yiren Zhang, Yuyang You, Keyang Hu, Kang Chen, Yu Chen. Turning Linux into a High-Performance Library OS with Flux. SOSP 2026.

    Paper Code

    • Flux turns Linux into a high-performance library OS while keeping mature Linux subsystems, including schedulers and file systems, largely unchanged. Unmodified applications invoke Linux system calls through direct function calls, while DPDK, SPDK, and a fast network path provide kernel-bypass I/O.

    • Flux combines single-level scheduling, Intel User Interrupts, optimized system-call paths, and a host-assisted process abstraction. It supports page-table-isolated processes and copy-on-write fork without sacrificing direct-call system calls.

    • Flux passes 92.2% of the selected Linux Test Project cases, improves ext4 Filebench throughput by up to 47% over Linux, comes within 4% of Junction’s Memcached throughput under a 100-microsecond p99 latency target, and improves eight-worker Nginx throughput by up to 18.5x over Linux under the same target.

  • Yuekai Jia*, Kaifu Tian*, Yuyang You, Yu Chen, Kang Chen. Skyloft: A General High-Efficient Scheduling Framework in User Space. SOSP 2024.

    Paper

    • Skyloft leverages user interrupts to support kernel-bypassing preemption, and is the first general user-space scheduling framework for multi-application workloads with microsecond-scale preemptive scheduling.

    • Skyloft introduces a scheduler development paradigm that builds diverse schedulers with a small set of common operations; policies that previously required heavy customization can now be implemented in only a few hundred lines.

    • Under both synthetic and real workloads on mainstream applications, Skyloft delivers performance that is competitive with, and often better than, existing solutions.

Internship#

  • Beijing Institute of Open Source Chip, Intern, 2025.6 - 2025.8 System Cache stream prefetcher design and implementation.

Competition Awards#

  • 6th Loongson Cup, Team First Prize, 2022

    • Implemented an out-of-order superscalar CPU in SystemVerilog, supporting two-level register renaming, dual-fetch triple-issue, and early wake-up for cached loads.

    • Debugged CPU implementation with the Difftest framework.

    • Achieved a final frequency of 120 MHz, capable of running Linux and Busybox.