About Me (English Version)


中文版

Wang Pengbo

Tel/WeChat: 15353506830

QQ: 1073693127

E-mail: pengbo_wang@foxmail.com

​ pengbo_wang@outlook.com

Current City: Suzhou, Jiangsu, China

Education Experience

Bachelor: Northwestern Polytechnical University | School of Computer Science | Computer Science and Technology | 2019 - 2023

GPA (first six semesters): 3.767

Comprehensive Ranking: 10/298

CCF-CSP: 270 (Top 6.28%)

Major Courses: Computer Composition and System Architecture (94), Compiler Principles (98), Database Principles (94), Signal and Systems Analysis (94), Algorithm Design and Analysis (91), Digital Logic Design (91), Computer Network Principles (90)

Master-PhD Program: University of Science and Technology of China | School of Computer Science and Technology | Computer Science and Technology | 2023 - 2028

Research Areas: Intelligent Computing Center Networks, Communication Acceleration for Large Model Training, Cloud Networks

Supervisors: Hongli Xu, Gongming Zhao

Research Group: USTC Intelligent Network and System Group (USTC-INT)

Project Experience

Interconnection of Multi-Kubernetes Clusters under Customized VPCs

Digital China Information Service Company Ltd., Main Developer, 2023.10 - 2024.7

  • Deployed multiple Kubernetes clusters and divided multiple VPCs in each cluster for isolation, aiming to enable cross-cluster communication among VPCs in different clusters
  • Conducted technical research on mainstream multi-Kubernetes interconnection solutions, with in-depth analysis of their system architectures and implementation mechanisms
  • Improved the open-source Submariner solution and built a control plane to synchronize cross-cluster service information and support multi-cluster service discovery
  • Used Go and Kubebuilder to develop gateway containers in each VPC to connect VPCs with the physical network, and established tunnels among clusters for efficient cross-cluster access

Communication Acceleration for Large Model Training: PS-AllReduce Backend Integration and DPU Offload (Megatron-LM + BytePS)

China Mobile Cloud Competence Center, Technical Lead, 2024.10 - Present

  • Analyzed the communication patterns of AllReduce under different parallel dimensions during large model training, integrated the BytePS communication library into Megatron-LM, and replaced/adapted its data-parallel and tensor-parallel communication paths
  • Integrated the BytePS (PS-AllReduce) backend into the DP gradient synchronization path of Megatron-LM, triggering push/pull based on grad bucket readiness and completing in-place gradient write-back after synchronization
  • Optimized the RDMA communication path of BytePS, identified resource contention caused by reusing QPs for push/pull, and promoted an architectural redesign with independent QPs for Push/Pull, improving RDMA bandwidth utilization from about 60% to over 80%
  • Advanced deployment on the BlueField-3 DPU side by migrating the server aggregation service to the DPU (aarch64), adapting the runtime environment and RDMA path, and completing compute offloading

Research Experience

Cross-Cluster Service Discovery and Access Design in Multi-Kubernetes Clouds

Published at the CCF-A conference The ACM Web Conference 2025

  • Existing multi-Kubernetes cluster interconnection solutions suffer from limited performance and insufficient robustness
  • Proposed a cross-cluster service discovery model based on a proxy-cluster mapping algorithm to provide highly available service information synchronization for multi-cluster environments
  • Deployed multiple cross-cluster gateways for aggregation, and optimized cross-cluster traffic forwarding with XDP/eBPF acceleration

Project Experience

NPU Dancing Robot Research & Training Base, 2020/07 - 2022/07

Core Member. Mainly responsible for embedded development and ROS-based robot control tasks. Proficient in using Linux operating systems. Experienced in STM32 and FreeRTOS development, as well as ROS robot development.

Publications

  • X-ClusterLink: An Efficient Cross-Cluster Communication Framework in Multi-Kubernetes Clusters
    Pengbo Wang, Gongming Zhao, Yuantao Wu, Hongli Xu, Haibo Wang
    The ACM Web Conference 2025 (WWW 2025), CCF-A Conference

  • Multi-tenant Deployment with Anomaly Isolation in Public Clouds
    Baoqing Wang, Jiawei Liu, Gongming Zhao, Jing Wen, Hongli Xu, Pengbo Wang
    IEEE/ACM International Symposium on Quality of Service 2025 (IWQoS’ 25), CCF-B Conference, short paper

Honors & Awards

Personal Honors:

  • National Scholarship for Master’s Students | University of Science and Technology of China, 2024 - 2025
  • National Scholarship for Undergraduate Students | Northwestern Polytechnical University, 2021 - 2022
  • Outstanding University Student | Northwestern Polytechnical University, 2019 - 2022, three consecutive years
  • First-Class Scholarship (Top 5%) | Northwestern Polytechnical University, 2020 - 2022, two consecutive years
  • “Samsung” Special Scholarship | Northwestern Polytechnical University, 2020 - 2021
  • Outstanding Communist Youth League Member | Northwestern Polytechnical University, 2020 - 2023, three consecutive years

Academic Competitions:

  • National Champion | 2021 RoboCup China Open, Rescue Robot Group, Dexterous Manipulation Event, 2021/05
  • National Runner-up | 2021 RoboCup China Open, Rescue Robot Group, Overall Results, 2021/05
  • National Runner-up | 2021 China Robot Competition, Dexterous Manipulation Event, 2021/04
  • National Runner-up | 2021 China Robot Competition, Target Detection Event, 2021/04
  • National First Prize | 2021 4th China University Intelligent Robot Creativity Competition, 2021/08
  • National Third Prize | 2021 “Huawei Cup” National College Student Internet of Things Design Competition, 2021/08
  • National Third Prize | 2020 3rd National College Student Embedded Chip and System Design Competition, 2020/11
  • National Bronze Prize | 2022 3rd National College Student Algorithm Design and Programming Challenge, 2022/03
  • Provincial Excellent Project Completion | 2021 College Student Innovation and Entrepreneurship Project, 2021 - 2022

文章作者: shenzu
版权声明: 本博客所有文章除特別声明外,均采用 CC BY 4.0 许可协议。转载请注明来源 shenzu !
  目录