Zhongke Yushu Launches New DPU Chip and Software Platform
TapTechNews June 20, it is reported that the 2024 product launch event of Zhongke Yushu was held yesterday at the Zhongguancun Exhibition Center in Beijing. At the launch event, Zhongke Yushu's third-generation DPU chip, K2 Pro, was officially released. It is currently the first mass-produced full-function DPU computing chip in China, customized and optimized for future data centers and cloud-native environments.
K2 Pro enhances support for complex business, integrates multiple types of hardware offloading engines such as network offloading, flow table offloading, storage offloading, and RDMA network offloading. The performance of complex service grids is reduced from 400 microseconds to less than 30 microseconds. In DPU complex scenarios, energy consumption is reduced by 30% to achieve low power operation.
K2 Pro will be installed in six DPU card products of Zhongke Yushu's three major product series. TapTechNews summaries as follows:
For ultra-low latency networks: SWIFT-2200N, SWIFT-NDPP
For high throughput lossless networks: FLEXFLOW-2200T, FLEXFLOW-2100R
For software-defined networks: CONFLUX-2200P, CONFLUX-2200E
Zhongke Yushu announces that its self-developed software development platform HADOS has been upgraded to version 3.0. HADOS 3.0 is specifically designed for DPU optimization. The core code volume has exceeded 1.26 million lines, and the cumulative total code volume is nearly 10 million lines. It has 2765 APIs at different levels such as driver, computing, storage, networking, and security, and has rich and ready-to-use modules and functions.
HADOS is adapted to 8 CPU platforms and 10 major operating systems, becoming one of the DPU software platforms with the most complete adaptation, the most competitive, and the most actual deployment in China.
In addition, Zhongke Yushu also released a high-performance cloud base solution with data network as the core - Yuyun, adopting the IaaS on DPU technical route. The Yuyun open platform built in the Xinchuang Park integrates over 400 high-performance servers and deeply integrates CPU, GPU and DPU technologies.