魏星达

长聘教轨副教授

wxdwfc@sjtu.edu.cn

个人简介

  • 魏星达，上海交通大学长聘教轨副教授。主要研究方向为操作系统和分布式系统。在包括OSDI/SOSP、EuroSys、NSDI等会议上发表多篇论文。英文主页见https://ipads.se.sjtu.edu.cn/pub/members/xingda_wei。

学术经历

  • 2024.06 - 至今，上海交通大学并行与分布式系统研究所，长聘教轨副教授

  • 2021.09 - 2024.06，上海交通大学并行与分布式系统研究所，助理教授

  • 2017.09 - 2021.06， 上海交通大学并行与分布式系统研究所，博士生

  • 2015.06 - 2017.07，上海交通大学并行与分布式系统研究所，硕士生

  • 2011.09 - 2015.06，上海交通大学软件学院，本科生

  • 研究项目

    • 下一代数据中心AI计算系统

      • PhoenixOS@SOSP’25 （GPU 弹性+容错）: 业界首个系统级并行GPU checkpoint和restore系统，性能超nvidia cuda-checkpoint一个数量级。
      • BlitzScale@OSDI’25 （弹性大模型推理）: 目前最快的弹性大模型推理系统。
      • KVCache@ATC’25 （大模型原生数据存储）: 基于真实推理场景实现智能AI中间数据（KVCache）管理，提升KVCache缓存利用率。

    • 下一代云计算操作系统

      • MITOSIS@OSDI’23：使用基于RDMA的Remote fork加速Serverless computing的冷启动。
      • Dmerge@EuroSys’24 （最佳论文）：新型分布式操作系统抽象，实现无序列化+反序列化数据传输。

    • 软硬协同的新型数据中心数据存储系统。

      • DrTM@SOSP’15, ATC’17 & OSDI’18 ：利用远端内存直接访问（RDMA）和硬件事务性内存（HTM）构建轻量级的内存事务系统
      • RDPMA@ATC’21：将RDMA和非易失性内存（NVM）高效进行聚合。
      • XStore@OSDI’22：首次使用机器学习模型加速基于RDMA的分布式键值存储系统。

获奖情况

  • 2024 ACM EuroSys 最佳论文奖

  • 2021 ACM 中国优秀博士学位论文提名奖

  • 2021 Honorable Mention of the ACM SIGOPS Dennis M. Ritchie Award（优胜奖）

  • 2021 ACM Chinasys 优秀博士论文奖

  • 2020 华为奥林帕斯 先锋奖

  • 2018 微软亚洲学者

  • 2017 Intel奖学金

  • 2015 上海交通大学优秀毕业论文

  • 发表论文 (部分)

    • [SOSP’25] Xingda Wei, Zhuobin Huang，Tianle Sun, Yingyi Hao，Rong Chen，Mingcong Han，Jinyu Gu，Haibo Chen. PhoenixOS: Concurrent OS-level GPU Checkpoint and Restore with Validated Speculation. The 31st Symposium on Operating Systems Principles. (To appear)
    • [OSDI’25] Dingyan Zhang, Haotian Wang, Yang Liu, Xingda Wei, Yizhou Shan, Rong Chen, Haibo Chen. Fast and Live Model Auto Scaling with O(1) Host Caching.. The 19th USENIX Symposium on Operating Systems Design and Implementation.
    • [USENIX ATC’25] Jiahao Wang, Jinbo Han, Xingda Wei, Sijie Shen, Dingyan Zhang, Chenguang Fang, Rong Chen, Wenyuan Yu, Haibo Chen. KVCache Cache in the Wild: Characterizing and Optimizing KVCache Cache at a Large Cloud Provider. 2024 USENIX Annual Technical Conference.
    • [EuroSys’24] Fangming Lu, Xingda Wei, Zhuobin Huang, Rong Chen, Mingyu Wu, and Haibo Chen. Characterizing Off-path SmartNIC for Accelerating Distributed Applications. 17th USENIX Symposium on Operating Systems Design and Implementation, Boston, MA, US, July 2023.
    • [OSDI’23] Xingda Wei, Rongxin Cheng, Yuhan Yang, Rong Chen, and Haibo Chen. Serialization/Deserialization-free State Transfer in Serverless Workflows. The 19th ACM SIGOPS European Conference on Computer Systems, Athens, Greece, April 2024.
    • [OSDI’23] Xingda Wei, Fangming Lu, Tianxia Wang, Jinyu Gu, Yuhan Yang, Rong Chen, and Haibo Chen. No Provisioned Concurrency: Fast RDMA-codesigned Remote Fork for Serverless Computing. 17th USENIX Symposium on Operating Systems Design and Implementation, Boston, MA, US, July 2023.
    • [USENIX ATC’21] Xingda Wei, Xiating Xie, Rong Chen, Haibo Chen, Binyu Zang. Characterizing and Optimizing Remote Persistent Memory with RDMA and NVM. 2021 USENIX Annual Technical Conference, July 2021. [paper]
    • [NSDI’21] Xingda Wei, Rong Chen, Haibo Chen, Zhaoguo Wang, Zhenhan Gong, and Binyu Zang. Unifying Timestamp with Transaction Ordering for MVCC with Decentralized Scalar Timestamp. The 18th USENIX Symposium on Networked Systems Design and Implementation, Boston, MA, US, April 2021. [paper]
    • [OSDI’20] Xingda Wei, Rong Chen, and Haibo Chen. Fast RDMA-based Ordered Key-Value Store using Remote Learned Cache. 14th USENIX Symposium on Operating Systems Design and Implementation, Banff, Alberta, Canada, November 2020.
    • [OSDI’18] Xingda Wei, Zhiyuan Dong, Rong Chen, and Haibo Chen. Deconstructing RDMA-enabled Transaction Processing: Hybrid is Better! 13th USENIX Symposium on Operating Systems Design and Implementation, Carlsbad, CA, US, October 2018. [paper]
    • [USENIX ATC’17] Xingda Wei, Sijie Shen, Rong Chen, and Haibo Chen. Replication-driven Live Reconfiguration for Fast Distributed Transaction Processing. 2017 USENIX Annual Technical Conference, Santa Clara, CA, US, July 2017. [paper]
    • [SOSP’15] Xingda Wei, Jiaxin Shi, Yanzhe Chen, Rong Chen, and Haibo Chen. Fast In-memory Transaction Processing using RDMA and HTM. 25th ACM Symposium on Operating Systems Principles, Monterey, CA, USA, October 2015. [paper]