Shulou(Shulou.com)11/24 Report--

CTOnews.com news on December 31, in the "2022 Tianyi Digital Science and Technology Ecology Conference" sponsored by China Telecom, China Telecom officially released Tianyi Yun Zijin DPU. DPU (Data Processing Unit) is a special processor constructed with data as the center.

The traditional computing architecture with CPU as the core has been unable to meet the computing needs of various industries. In recent years, Moore's Law has slowed down, and the improvement of CPU performance is difficult to meet the growth rate of computing demand. Part of the host CPU computing power is used to host virtualized applications, which will consume more than 30% of computing power. At the same time, a wide variety of CPU chip architecture and operating systems lead to high adaptation difficulties and heavy workload. CTOnews.com learned that DPU can achieve operational separation of business and infrastructure, offload computing power such as network, storage and security, and greatly improve performance and support computing and storage separation.

China Telecom Tianyi Yun launched a completely self-developed Zijin DPU chip, and based on Zijin DPU, created a new generation of cloud computing architecture for the cloud, migrating computing, storage, network and secure computing power to a completely self-developed software-hardware combination system, and building independent and controllable technologies for different computing power and national production.

According to China Telecom, based on Tianyi Cloud's full self-research capability, the architecture has completed three major technological breakthroughs: virtualization full uninstallation, multi-service hardware acceleration and self-development of high-performance network protocols.

① Virtualization full Uninstall Tianyi Cloud Zijin architecture uninstalls all kinds of virtualization components originally running on server CPU to Zijin DPU to achieve zero loss of server virtualization and fully release computing power, which not only improves the sales efficiency of elastic CVM, but also supports elastic bare metal services.

② multi-service hardware acceleration has higher performance and lower latency. Compared with the traditional CPU processing method that relies on pure software, Zijin DPU improves network PPS performance by 100% through chip-level hardware acceleration, improves storage IOPS performance by 200%, and reduces network latency by 300%. The overall performance of the new architecture has been greatly improved.

③ provides high-performance network protocol Tianyi Yunzijin DPU to fully hardware the high-performance network data surface, NVMe-oF protocol, EC error correction code and other functions. The self-developed congestion control algorithm achieves accurate flow control through end-to-end network cooperative active control, achieves ultra-low latency through low queue depth, greatly improves the communication efficiency of distributed computing, and has the advantage of large-scale deployment. At the same time, Zijin DPU also realizes the physical isolation of the host CPU environment from the virtualized environment, supports "one cloud with multiple cores", and implements "plug and play" for servers with different CPU chip architectures.

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.