About Me

I am an software engineer with over four years of experience developing high performance networking technologies and open source software. As a Senior Software Engineer at Intel, I have established myself as an expert in accelerating packet processing workloads and network virtualization. Through extensive contributions to the Data Plane Development Kit (DPDK) open source project, I have demonstrated expertise optimizing network performance and unlocking the potential of infrastructures. I pioneered the development of DPDK’s Asynchronous Vhost and Mirroring Libraries, enabling substantial performance and scalability gains for Virtio/Vhost applications. I also architected dma-perf, DPDK’s first performance analysis toolkit for offloading data-intensive workloads. Beyond technical achievements, I have contributed to guiding open source security practices impacting thousands of users worldwide. I spearhead community-wide efforts coordinating cross-functional teams to diagnose and remediate critical vulnerabilities within tight deadlines. Through dedication and security acumen, I have aimed to uphold the reliability and robustness of fundamental open source technologies. With a passion for meaningful work and hands-on experience driving advancements in networking, I aspire to leverage my skills to create technologies that make the world a better place.

Experiences

Network Software Development Engineer

2019 - Present
Intel Asia-Pacific Research & Development Ltd. Shanghai

Through sustained participation and contributions to the DPDK project over the years, I have developed some understanding and experience in fields like high performance networking, hardware acceleration and virtualization. By spearheading development of the Asynchronous Vhost and Mirroring Libraries, the DMA performance testing application dma-perf, and optimizations to the DMAdev framework, I have demonstrated my technical abilities to leverage DMA and achieve substantial performance gains in environments ranging from vhost virtualization to Hardware integration. My work continues driving innovations in open source technologies unlocking the potential of networking infrastructure to meet rapidly increasing workload demands.

  • DPDK Asynchronous Vhost Library Development:

    • Architected and implemented an asynchronous vhost library for DPDK utilizing Intel® Data Streaming Accelerator (DSA) to accelerate Tx/Rx data plane operations of packed virtqueues.

    • Conducted extensive performance analysis and optimization on virtio/vhost implementations, achieving substantial performance improvements through elimination of redundant data copies and exploitation of DSA’s advanced DMA capabilities. The library was accepted by the DPDK community and integrated into the mainstream DPDK release.

    • Harnessed the power of DMA engine and asynchronous offload model to enable a high performance zero-copy data path from guests to wire. The optimized vhost implementation demonstrated over 50% throughput gains and 30% CPU utilization reduction compared to the previous architecture.

    • Collaborated with Intel’s DSA engineering team on feature requirement analysis, optimization techniques and product roadmap feedback based on use case scenarios. The productive collaboration resulted in a number of key hardware and software updates to maximize the DPDK Vhost/DSA synergy.

    • Spearheaded the proposal and development of DPDK’s first fully asynchronous vhost library. The new design provides a non-blocking, event-driven framework to overlap DSA data transfers with concurrent vhost operations, resulting in significant performance and scalability improvements for virtio/vhost applications.

  • DPDK Asynchronous Vhost Library Development:

    • Architected and built an asynchronous port mirroring library for DPDK utilizing Intel® Data Streaming Accelerator (DSA) to accelerate real-time traffic replication between vhost port queues.

    • Implemented high-performance DMA-based data plane APIs to copy Tx/Rx traffic from monitored to listening ports. The optimized mirroring implementation demonstrated over 150% higher monitoring throughput and 30% reduced CPU utilization compared to a CPU-only solution.

    • Spearheaded the design and development of DPDK’s first fully asynchronous port mirroring library. The non-blocking framework overlaps DSA mirroring operations with concurrent port traffic to minimize performance impact, enabling lossless, line-rate traffic replication at maximum port bandwidth.

    • Presented the DPDK Port Mirroring solution at industry conferences, demonstrating its significant value for network monitoring, security and visibility applications. The library has gained wide adoption in the DPDK community.

  • Designed and Developed DPDK Intel DMAdev Driver

    • Implemented and benchmarked three DSA features: Memory Compare, CRC Generation, and Delta Record Creation. Evaluated performance profiles to determine optimal use cases.

    • Optimized hash table lookups by offloading compute-intensive memory comparison and CRC functions to the DSA. Achieved substantial performance gains, enhancing throughput and scalability of DPDK applications.

    • Utilized DSAs to accelerate pattern matching and cyclic redundancy checks for hash tables containing millions of entries.

    • Demonstrated strong technical and analytical abilities to optimize data access patterns and accelerate system performance through co-processor utilization.

    • Showcased software engineering and optimization skills that would allow me to make impactful contributions as a network software engineer.

    • Gained valuable expertise in benchmarking, performance analysis, and hardware-software co-design.

  • Design and Development of a Performance Analysis Toolkit for DMA

    • Designed and developed a DMA performance testing application (DMA-perf) based on the DPDK DMAdev framework.

    • Utilized configuration files to facilitate setup of various performance testing scenarios. Recorded results in designated output files for analysis.

    • Supported multiple test scenarios to evaluate DMA device hardware performance under diverse conditions, including cache usage and workload variations.

    • Focused on determining optimal methods for offloading memory-intensive workloads to DMA engines for maximum efficiency.

    • Showcased strong technical skills in framing and executing rigorous performance evaluations, benchmarks, and tests. Able to build customized tools and applications to analyze system capabilities.

    • Demonstrated expertise in DMA programming, performance optimization, and hardware-software co-design.

    • Made significant contributions to enhancing utilization, optimization, and applications of DMA through targeted performance analysis.

    • Gained valuable experience developing DPDK-based tools and applications to assess and improve the performance of complex network systems.

  • DPDK Community Security Team Member

    • Serve as a core member of the DPDK open source community security team, overseeing vulnerability management and risk mitigation strategies that impact thousands of users worldwide.

    • Demonstrate influential leadership, technical expertise, and project management skills guiding the community’s security practices and responses. Develop processes that became model security frameworks adopted by other open source projects.

    • Make impactful contributions upholding responsibility for the security and integrity of open source technologies fundamental to the networking industry. Backport multiple security fixes to stable releases deployed across enterprise production environments.

    • Build a record of effective leadership and achievement within the open source community through dedication, technical acumen, collaborative ability, and communication skills. Develop a strong reputation as a security subject matter expert.

  • DPDK Community DMA-perf Application Maintainer

    • Serve as a maintainer of the DPDK DMA-perf application, contributing performance enhancements utilized by hundreds of users worldwide.

    • Demonstrate sound technical expertise and innovative ideas optimizing performance and efficiency. Maintain the application’s code quality and reliability through ongoing testing and bug fixes in new releases.

    • Make impactful contributions furthering the performance and usability of an application integral to optimizing networking performance. Backport critical bug fixes and enhancements to stable releases ensuring broad compatibility.

    • Build a reputation for technical excellence and commitment within the DPDK community through continuous involvement, code reviews, and responsiveness to bug reports and requests for assistance. Establish a strong track record as an authority on DMA performance optimization techniques.

Software Development Intern

2018 - 2019
Intel Asia-Pacific Research & Development Ltd. Shanghai
  • Conduct in-depth analysis of network performance issues and bottlenecks within container and virtualized environments. diagnose complex problems spanning application scenarios, protocols, and architectures.

  • Design and implement experimental frameworks to rigorously evaluate network performance of technologies like Docker. Analyze results and obtain data-driven insights into optimization opportunities.

  • Make enhancements to open source software such as gRPC to improve performance and functionality. Extend systems and tools leveraging technical skills in C/C++ and debugging.

  • Perform code reviews, architectural evaluations, and proofs of concept for technologies like Envoy, service mesh, and sidecar patterns. Identify potential for performance improvements and accelerated data plane processing through DPDK.

  • Research and review the latest publications and advancements in fields of network virtualization, protocol optimization, cloud native architectures, and more. Maintain cutting-edge knowledge of trends and developments in networking technologies.

  • Demonstrate fluency across the networking stack, virtualization technologies, and distributed system architectures. Apply technical depth and hands-on skills to make meaningful performance optimizations and enhancements.

  • Stay up-to-date with innovations in open source software, networking models, and distributed computing through continuous learning and participation in the technical community. Rapidly assimilate new technologies and architectures.

Projects

Visual Studio Code debug arguments generator - This application converts command line arguments into launch.json configuration parameters for Visual Studio Code debugging. The launch.json file defines various debug configurations to control the debugger in VS Code. However, manually editing JSON files can be tedious and error-prone. This application automates the process by taking command line arguments from the user and translating them into the JSON properties and values to populate the launch.json.

OSS Contributions

Data Plane Development Kit - Data Plane Development Kit (DPDK) consists of libraries to accelerate packet processing workloads running on a wide variety of CPU architectures.

Publications

  • Accelerating DPDK Vhost Interface via Intel Data Streaming Accelerator
  • Hu Jiayu, Jiang Cheng
    Intel Network Technical Summit, 2023
  • Intel® Data Streaming Accelerator (DSA) - Accelerating DPDK Vhost Technology Guide
  • Hu Jiayu, Jiang Cheng, Ding Xu, He Xingguang
    Intel Technology Guide, 2023
  • DPDK社区安全问题处理流程介绍
  • Jiang Cheng
    DPDK与SPDK开源社区, 2022
  • 异步模式下的 Vhost Packed Ring 设计介绍
  • Jiang Cheng
    DPDK与SPDK开源社区, 2022
  • Fabrication and analysis of large-area three-dimensional photonic crystals
  • Jiang Cheng, Xue Yanling, Wang Rui
    Journal of Optoelectronics and Laser, 2018

    Skills & Proficiency

    C

    Python/Matlab