Hello, this is Hanyu.

I am Hanyu Zhao (赵汉宇), an engineer and researcher at the Platform for Artificial Intelligence (PAI) team of Alibaba Cloud, where we investigate and build distributed systems for modern AI workloads to support Alibaba's and other customers' large-scale services. I obtained my PhD degree from Peking University in 2021, supervised by Prof. Yafei Dai and Prof. Zhi Yang. I was an intern at the Systems Research Group of Microsoft Research Asia from 2017 to 2020, where I had a great time working with and learning systems research from many brilliant people.
I am interested in distributed systems and systems for AI. My recent focus is on cluster-scale AI training and serving, with a particular attention to the classic distributed systems challenges in this new context, such as cluster scheduling, fault tolerance, and data storage. I also studied distributed consensus and consistency protocols during my PhD.
I like systems research motivated by real-world problems, driven by deep insights into the problem, devising principled, systematic, and elegant solutions with solid and practical implementations. Difficult as it is, I will keep pursuing it :)
Here is my (outdated) Curriculum Vitae (中文简历).

Selected Publications

See the full list at Google Scholar.

  • Rubick: Exploiting Job Reconfigurability for Deep Learning Cluster Scheduling
    Xinyi Zhang, Hanyu Zhao, Wencong Xiao, Xianyan Jia, Fei Xu, Yong Li, Wei Lin, Fangming Liu
    arXiv preprint (2408.08586)
    [Paper]
  • Llumnix: Dynamic Scheduling for Large Language Model Serving
    Biao Sun*, Ziming Huang*, Hanyu Zhao*, Wencong Xiao, Xinyi Zhang, Yong Li, Wei Lin (*co-first authors)
    18th USENIX Symposium on Operating Systems Design and Implementation (OSDI '24)
    [Paper/Slides/Talk] [Code]
  • EasyScale: Elastic Training with Consistent Accuracy and Improved Utilization on GPUs
    Mingzhen Li, Wencong Xiao, Hailong Yang, Biao Sun, Hanyu Zhao, Shiru Ren, Zhongzhi Luan, Xianyan Jia, Yi Liu, Yong Li, Wei Lin, Depei Qian
    International Conference for High Performance Computing, Networking, Storage, and Analysis 2023 (SC '23)
  • GoldMiner: Elastic Scaling of Training Data Pre-Processing Pipelines for Deep Learning
    Hanyu Zhao, Zhi Yang, Yu Cheng, Chao Tian, Shiru Ren, Wencong Xiao, Man Yuan, Langshi Chen, Kaibo Liu, Yang Zhang, Yong Li, Wei Lin
    2023 ACM International Conference on Management of Data (SIGMOD '23)
    [Paper/Talk]
  • SiloD: A Co-design of Caching and Scheduling for Deep Learning Clusters
    Hanyu Zhao*, Zhenhua Han*, Zhi Yang, Quanlu Zhang, Mingxia Li, Fan Yang, Qianxi Zhang, Binyang Li, Yuqing Yang, Lili Qiu, Lintao Zhang, Lidong Zhou (*co-first authors)
    18th European Conference on Computer Systems (EuroSys '23)
    [Paper]
  • HiveD: Sharing a GPU Cluster for Deep Learning with Guarantees
    Hanyu Zhao*, Zhenhua Han*, Zhi Yang, Quanlu Zhang, Fan Yang, Lidong Zhou, Mao Yang, Francis C.M. Lau, Yuqi Wang, Yifan Xiong, Bin Wang (*co-first authors)
    14th USENIX Symposium on Operating Systems Design and Implementation (OSDI '20)
    [Paper/Slides/Talk] [Code]
  • SDPaxos: Building Efficient Semi-Decentralized Geo-replicated State Machines
    Hanyu Zhao, Quanlu Zhang, Zhi Yang, Ming Wu, Yafei Dai
    ACM Symposium on Cloud Computing 2018 (SoCC '18)
    [Paper] [Slides] [Talk] [Code]
  • Gandiva: Introspective Cluster Scheduling for Deep Learning
    Wencong Xiao, Romil Bhardwaj, Ramachandran Ramjee, Muthian Sivathanu, Nipun Kwatra, Zhenhua Han, Pratyush Patel, Xuan Peng, Hanyu Zhao, Quanlu Zhang, Fan Yang, Lidong Zhou
    13th USENIX Symposium on Operating Systems Design and Implementation (OSDI '18)

Life

In my leisure time, I do various types of workout (gym/jogging/climbing/...), through which I managed to lose 25+kg of weight and said goodbye to 10+ years of being a fatty. I also enjoy taking photos (the top banner is a recent "Shot on iPhone"), singing songs, and playing poker (especially Hold'em!). Join me if you're interested.
I am a super fan of cars. Putting my hands on the steering wheel always makes me feel relieved. I'm currently driving a Volvo V60 wagon, which is pretty rare in China. Also dreaming of having one of Porsche 911 or BMW M Power some day!
I used to be a member of the Debate Team of Wuhan University, one of the best Mandarin debate teams in the world :)