Introduction to the Business of Computer Network & Protocol Laboratory & Distributed and Parallel Software Laboratory

Computer Network & Protocol Laboratory

Huawei’s Computer Network & Protocol Laboratory is a subsidiary of the Distributed and Parallel Software Laboratory of the Central Software Institute of the 2012 Laboratory, with locations in Beijing, Shanghai, Hangzhou, Shenzhen, and Tel Aviv, Israel.

Vision: Rooted in laying the foundation stone, innovation leads the future of distributed communication

Positioning: Huawei’s software engine in the field of computer network and protocol technology, covering theoretical breakthroughs, technological inventions, technological innovations, and quality delivery. Standing at the forefront of this technical field, we research and break through world-class technical problems in computing native networks and wide-area network deterministic communication, build an industry-leading full stack of distributed communication, and work with ICT, terminal, cloud, intelligent car and other main product teams to build differentiated communication competitiveness, gradually grow the industrial ecosystem, and help business success.

Team: A high-level innovation team composed of top industry-leading experts, genius youngsters, PhDs, and engineers mixed special forces, overseas legions. The technical research results are significant. Since 2018, 5 papers have been accepted by SIGCOMM, the top global network communication conference; and key technologies have been selected into Huawei’s top 10 inventions for three consecutive sessions.

【Key Research Directions of the Laboratory】

1. Data Center Network Communication Technology: Aimed at the ultimate pursuit of “low-latency communication” for large-scale heterogeneous parallel computing systems, define the software standards for the interconnection of the next generation of computing native networks, innovate the “memory semantic” distributed software architecture, release the hardware characteristics of the integrated communication bus, enable the overall computing power of the distributed system to grow linearly, and build a new ecosystem of high-performance communication middleware.

• Industry: The world’s top chip giants and internet cloud service giants are increasing their technological innovation in high-performance computing networks (Intel PCIe, NVIDIA NVLink, Mellanox infiniband and UCX, AWS SRD/EFA, Alibaba Cloud’s HPCC/etc.)
• Typical Results: eVS (10 million level PPS elastic virtual switch), CurreNET (large-scale low-latency RDMA network), and the company-level strategic project Unified Bus (next-generation data center network bus with memory semantics) and UMDK (high-performance memory semantic communication suite) jointly built with Hisilicon and Zhongying departments.

2. Wide Area Network Communication Technology: Aimed at the deterministic communication guarantee requirements put forward by emerging applications such as large-scale real-time interactive live broadcasting, RTC real-time audio and video communication, Cloud XR, V2X, and remote monitoring and control in the industrial field, innovate the interconnection system architecture and protocol of application-driven real-time networks, enable and release the potential of new infrastructure such as 5G slicing networks, and create cost-effective quasi-deterministic communication services.

• Industry: Internet giants are building their own global acceleration networks (for example: Google Cloud Global Acceleration Network, AWS Global Acceleration Network, Microsoft Azure Global Network, Facebook Global Network, Alibaba Cloud Global Network, Tencent Cloud Global Network, etc.), and a large number of CPaaS (Communication Platform as a Service) manufacturers represented by Twilio, Agora, and Ronglian Cloud are emerging and growing rapidly.
• Typical Results: Link Turbo all-network aggregation acceleration technology jointly built with the terminal team, Network Kit/hQUIC Kit/RTN Service jointly built with the HMS team; SparkRTC jointly built with the cloud service team.

3. Wireless Local Area Network Communication Technology: Aimed at the stable low-latency and high-bandwidth communication needs of seamless collaboration between 1+8+N devices in all scenarios of terminal intelligence, innovate extremely simple communication network and protocol technology, build the core engine of HarmonyOS and HMS distributed communication, and enable superior terminal distributed quality experience.

• Industry: Intelligent terminal giants are increasing their technological layout for device interconnection (for example: Apple’s own ecosystem HomeKit/Airdrop/Airplay, Samsung’s SmartThing, Google’s Matter, and the mutual transmission alliance jointly established by Xiaomi/OPPO/VIVO, etc.)
• Typical Results: Huawei Share, mobile phone clone, HarmonyOS distributed soft bus jointly built with the terminal team, and the new experience of cross-end distributed games jointly innovated; Nearby Service jointly built with the HMS team.

Related Technologies: Data center network, wide area network, wireless local area network, vehicular network, industrial internet network, high-performance communication middleware (UCX/MPI/RPC, etc.), data-driven traffic engineering, network parameter automatic tuning, real-time audio and video communication, network fault localization, formal verification, high-performance communication protocol (RDMA, QUIC, TCP/IP, etc.), mainstream programming languages (C, C++/Java/Python, etc.).

Distributed and Parallel Software Laboratory

The Distributed and Parallel Software Laboratory belongs to the Central Software Institute of the 2012 Laboratory, and is divided into the Distributed Parallel Computing Laboratory, Distributed Data Laboratory, Computer Network & Protocol Laboratory, Poisson Laboratory, Riemann Laboratory, etc. The laboratory is located in Hangzhou, Shenzhen, Beijing, Shanghai, Wuhan, Toronto, Tel Aviv, and other regions, and each subordinate laboratory is basically distributed in multiple regions, so you can basically choose the city of life freely.

The director of the laboratory is Dr. Tan Kun, a well-known expert in the field of distributed systems and networks, and the laboratory also has many senior experts, genius youngsters, senior engineers, etc. from the industry and academia.

The laboratory’s research results are significant, including MindSpore, MindData, Yuanrong Serverless Framework, Tongtu Network Protocol Stack, Poisson Search, Riemann Map, etc.

Here’s how I feel about the business of the Distributed and Parallel Software Laboratory, which doesn’t represent the official explanation, just a soldier’s thinking:

• Datacenter As a Computer: The Unified Bus provides high-performance interconnection for data centers, making the interconnection of data centers as simple as a computer; Yuanrong Serverless provides a new abstraction for distributed system programming, and MindSpore provides a new abstraction for AI programming, making programming in data centers as simple as a computer, enabling “single-machine thinking, distributed parallel execution” (When I introduce MindSpore, I often say it’s PyTorch’s skin + TensorFlow’s heart); Based on this, a new type of data analysis middleware that integrates AI, scientific computing, and big data processing capabilities is built, allowing data scientists to achieve interactive analysis of big data.
• Next-generation Search Advertising, Digital Maps: Based on the powerful computing power of “edge-cloud as a computer”, search is extended from perception to cognition, making human-computer interaction more natural, deeply integrating search and recommendation, advertising, and still achieving good recommendation results under increasingly strict privacy protection regulations; Relying on key technologies of high-precision map construction, positioning and navigation, build the digital foundation for the future metaverse.

Welcome to join the Distributed and Parallel Software Laboratory and the Computer Network and Protocol Laboratory. You can add my WeChat bojustc, or use the other contact methods listed on my personal page.