It is called the third main chip of the data center. Why does DPU
Time:2022-10-22
Views:1620
Source: Huaqiang Electronic Network Author: Well Core Microenterprise
From October 2021 to now, the Institute of Computing of the Chinese Academy of Sciences has released the White Paper on Special Data Processor (DPU) Technology and the Evaluation Method and Implementation of the Performance Benchmark of Special Data Processor (DPU) for the whole industry. The two white papers explain more comprehensively why DPU is defined as a disruptive technology for the development of data centers.
In the white paper on DPU performance benchmark evaluation, the evaluation and implementation methods of DPU in the four areas of network, storage, computing and information security are described. These are also the four areas where DPU releases its potential. At the same time, they are the key links of the data center. Therefore, DPU, CPU and GPU are called the troika of the computing economy era, and DPU has become the "third main chip" in the data center field. In addition, intelligent driving, data communication and network security are also the target application fields of DPU. According to the estimation of CCID, the global DPU market size will exceed 10 billion dollars from 2023, and the subsequent annual growth rate will exceed 50%. In the white paper on DPU performance benchmark evaluation, the evaluation and implementation methods of DPU in the four areas of network, storage, computing and information security are described. These are also the four areas where DPU releases its potential. At the same time, they are the key links of the data center. Therefore, DPU, CPU and GPU are called the troika of the computing economy era, and DPU has become the "third main chip" in the data center field. In addition, intelligent driving, data communication and network security are also the target application fields of DPU. According to the estimation of CCID, the global DPU market size will exceed 10 billion dollars from 2023, and the subsequent annual growth rate will exceed 50%.
Source: CCID
This is why Nvidia CEO Huang Renxun stressed that "this is just the beginning" when he released the company‘s first DPU product At present, not only industrial giants such as Nvidia are closely following the DPU track, but also start-ups such as Wellcore Microelectronics have come to the fore, showing extraordinary technical strength, and their products have been recognized by the industry.
What is a DPU?
If you want to know why the DPU is hot, you must first figure out what the DPU is?
The full English name of DPU is Data Processing Unit, which translates directly into data processing unit and is also called data center processor by some practitioners. DPU is data centric and its core role is to accelerate data processing, helping the system solve the load that CPU processing efficiency is low and GPU cannot handle.
According to the DPU technical white paper, the data processed by the DPU is different from the "signals" processed by traditional DSP, baseband processors, and the graphics and image data processed by the GPU. The data processed by the DPU is various kinds of information closer to the underlying architecture and digitalization, especially all kinds of sequential and structured data, such as large structured tables, data packets in network streams, massive texts, and so on.
DPU architecture reference design
Source: DPU Technical White Paper
DPU has great potential
It is not difficult to find that DPU is proposed to solve the bottleneck of data center development as the first starting point, and it is the biggest beneficiary of the data center using servers to move towards heterogeneous computing. With the vigorous development of the data center market, DPU has great potential in this field.
Taking the Chinese market as an example, the first wave of data centers started in 2014-2015, mainly serving the rapid development of the Internet at that time. At that time, there were many data exchanges in China. From the perspective of specific characteristics, the first wave of data center is computing centered. However, the way of using CPU as the main computing platform soon became a bottleneck, and 20% - 30% of computing resources were wasted in CPU resource management, which brought a sense of powerlessness to improve computing power.
The second wave of data centers began in 2020, which is known as the "first year of DPU". The most obvious change is that data centers no longer emphasize the concept of computing center and computing power center, but return to the nature of data center. In this process, the types of cloud and application algorithms are becoming more and more diverse. Of course, improving the efficiency of data processing is still the main goal, but it is also a general direction to reduce the burden on the CPU to enhance the system scheduling capability.
From "computing centric" to "data centric" is a huge conceptual change, and DPU has been able to stand out from many coprocessors. As a dedicated engine for data processing, the increasingly rich product types of DPU in the future will enable the CPU to return to its field of expertise - overall planning and task scheduling, and release the CPU‘s computing power to upper applications. DPU will become a new system gateway, storage portal, algorithm acceleration sandbox, and security engine.
As can be seen from the figure below, DPU is ubiquitous in the data center framework.
Traditional cloud computing infrastructure and data oriented? In the data of? Architecture pair?
Source: DPU Technical White Paper
Of course, DPU also needs to overcome some challenges in order to realize its potential in the data center field. First of all, at the hardware level, the DPU product categories need to be rich enough. As mentioned above, the CPU needs to be fully unloaded in the four dimensions of network, storage, computing and information security. At the same time, DPU products from different manufacturers need to follow common standards and reflect the total cost of ownership advantage in the overall scheme; At the software level, the current application of CPU and GPU in the data center has a standard software framework, which realizes the decoupling of software and hardware. After DPU is joined, this is also a long-term and challenging work.
As mentioned above, the terminal application scenario of DPU is not only in the data center, but also in the fields of intelligent driving, data communication, network security, etc. DPU has broad application prospects.
For example, in the field of communication, the DPU based "Data Transfer" is implemented? Network function virtualization can be applied to 5G edge computing UPF to improve the high-speed and large bandwidth capabilities of 5G communications in vertical industries; In the field of network security, DPU can help build hardware trust roots and achieve isolation by completely uninstalling? Network virtualization, isolated from the host operating system, protects users from hackers; In the field of automatic driving, various virtualization functions provided by DPU can better help developers realize "software defined cars".
Domestic DPU helps China‘s "new infrastructure"
DPU has opened a big industrialization trend, and all fields in China‘s "new infrastructure" strategy will take the lead in benefiting from it and accelerating the digital transformation. At the same time, "new infrastructure" and "data from the east to the west" will also bring strong demand in terms of data and computing power. In addition to the country‘s grand layout in 5G and automatic driving industries, the future development prospects of domestic DPUs are bright.
Seizing the opportunities of the DPU industry in China, Wellcore Microelectronics (Tianjin) Co., Ltd. (hereinafter referred to as Wellcore Microelectronics), although still a young start-up, has reached the domestic leading level in terms of DPU technology research and development and product commercial progress.
Well Core Micro was established in Tianjin Economic Development Zone in 2020, focusing on new computing, new network, new security and other innovative tracks, and bringing leading DPU products and technical services to target customers in key industries such as the party, government and military, finance, energy, transportation, power, and telecommunications.
According to the DPU technical white paper, the DPU architecture has three core components, namely, the control plane, the IO subsystem and the data plane. In terms of IO subsystem, Well Core Micro provides original technologies such as RapidIO embedded system interconnection and software definition interconnection and switching, and has successfully developed system IO products such as RapidIO switching chip NRS1800, software definition interconnection switching chip SDI3210, endogenous security switching chip ESW5610, and bridge chip PRB0400.
In the data plane, Wellcore Microenterprise has released a network processing intelligent chip integrated with AI engine, Wellcore 2820, which is mainly oriented to cloud computing and data center network applications and related task unloading, and provides intelligent network monitoring and management capabilities. The processing of network packets is the first power generation scenario of the DPU. The CPU is unloaded in the "network+computing power" mode. Therefore, the DPU is often regarded as an upgraded version of SmartNIC (Smart Network Interface), but we all know that the potential of the DPU does not stop there.
The Wellcore 2820 has realized the message analysis of Layer 2-7 and the acceleration ability to deal with cloud computing, industrial Internet and other key applications. At the same time, the chip is scalable, able to cope with rapidly changing big data processing and delay sensitive application requirements, and has the ability to rapidly deploy.
Well core 2820 supports three application scenarios:
Standard network card mode In this mode, the chip works in a standard network card mode, and can be configured as a 2 * 100GE standard network card, a 4 * 10GE network card, a 2 * 25GE network card, or a 2 * 10GE network card. It supports up to 16 * PCIE4.0 and SR-IOV data transmission of multiple virtual machines. The software accesses the network interface card through the standard linux network card driver.
Intelligent network card mode In this mode, the CPU and AI engine participate in data processing, and the chip works in the intelligent unloading mode. The data message is sent through two paths, one is the control path, and the related protocol message is forwarded to the CPU core for processing of various protocol control surfaces such as RDMA and OVS; One is the data path, which is used to process the encapsulation/de encapsulation of messages, data processing acceleration, etc. AI engine extracts and analyzes input data in real time, detects and identifies whether data traffic has malicious attacks, and reports and alerts.
Independent master mode Under this mode, the well core micro network processing intelligent chip can work independently and become a network master CPU chip with AI engine. PCIE works in RC mode and can be connected with various peripherals, such as 24 port/48 port gigabit switching chip, SRIO switching chip, etc. The CPU runs an independent Linux operating system and network protocol stack to realize the main control function of the switch.
The launch of this dual 100G network processing smart chip represents a key step for local enterprises in the DPU field. Later, Wellcore Micro will continue to focus on the three strategic directions of software definition interconnection, endogenous security and brain like computing, promote the rapid iteration of DPU and other products, and contribute innovation to the promotion of the national "new infrastructure" strategy.
|
Disclaimer: This article is transferred from other platforms and does not represent the views and positions of this site. If there is any infringement or objection, please contact us to delete it. thank you! |
Business consulting
13823761625
Mail me