On August 10, the 2023 Open Computing China Community Technology Summit (OCP China Day 2023) was held in Beijing. In the age of wisdom, computing diversification, application diversification, and technological complexity are driving a new round of transformation in the data center. The open source and open community has become an important force to promote continuous innovation in the data center. Through global collaboration and innovation, we will work together to solve the problem of data center infrastructure. Major issues such as iteration and sustainable development.
Since 2019, OCP China Day has become an annual event where industry leaders discuss open computing technology and innovation. This summit is sponsored by the open computing community OCP and undertaken by Inspur Information. With the theme of "Open Momentum: Intelligent, Scalable and Sustainable", it focuses on data center infrastructure innovation and sustainable development, open computing innovation and ecological construction, community Collaborative innovation and other issues, from community leaders and technical masters including OCP Foundation, Microsoft, Inspur Information, Samsung, Amou Technology, Douyin, Jingfeng Mingyuan, Alibaba Cloud, JD Cloud, Meituan, Suiyuan Technology, etc. , industry experts, and more than a thousand IT engineers and data center practitioners participated in the conference.
Five major organizations gather , open source and openness break the boundaries of innovation
The deep integration of the digital economy and the real economy, as well as the digital world and the physical world, is driving data centers to accelerate innovation. However, increasingly complex technologies and new applications make it increasingly difficult for information technology service providers and enterprises to control the comprehensive innovation of data centers. Various open source and open communities with global collaboration as the core purpose have emerged as the times require, and have gradually become an important force for data center innovation.
George Tchaparian, CEO of the OCP Foundation, said: "The core of the spirit of open source and openness is to advocate freedom, equality, sharing and co-governance for global collaborative innovation. Innovation. As digital technologies become more and more complex, the progress of the community not only depends on a friendly and smooth collaborative development process and a complete community governance mechanism, but also needs to strengthen cross-community cooperation and sharing, and continuously expand the boundaries of innovation."
In consideration of the coordinated development of data center software and hardware, cross-community cooperation has been increasingly valued by open source community leaders and members. This summit brought together community leaders from five open communities and organizations, OCP, OCTC, OpenPOWER Foundation, Dragon Lizard Community, and CXL Technology Alliance, to discuss how to build a data center infrastructure with highly coordinated software and hardware from different perspectives.
Ma Tao, chairman of the Dragon Lizard Community, believes that the rise of open computing has greatly accelerated the speed of hardware innovation, and promoted the integration of software-defined functions and the improvement of efficiency. Dragon Lizard is committed to creating a next-generation operating system for the cloud era, driving R&D and innovation centered on cloud scenarios, which can break through the problem of insufficient driving force for the development of the computing industry after Moore's Law expires, and efficiently promote the highly coordinated development of software and hardware ecology.
Meow Yee, chairman of the OpenPOWER Foundation, said that open computing makes hardware agile innovation possible, and its value has been repeatedly verified in the past ten years. The OpenPOWER Foundation is committed to fully exploiting the leading technical advantages of OpenPOWER-related assets such as Power ISA, Open firmware, and Power cores, actively uniting member organizations all over the world, and building a more open, collaborative and win-win ecosystem, which not only accelerates innovation in the open hardware industry, but also At the same time support and accelerate the innovation of open source software.
Chen Hai, secretary-general of the OCTC Open Standards Working Committee, said that convergence will be the basic direction for the development of next-generation data centers. This kind of integration requires not only a high degree of coordination between software and hardware, but also the integration and innovation of open computing innovations and general industry application requirements. OCTC is committed to innovation-oriented architecture integration, building world-leading norms and standards through open co-construction and collaborative innovation, accelerating the implementation of open computing in the entire industry, and promoting the high-quality development of the ICT industry.
Left-handed technology and right-handed standard, open source and open source should pay equal attention to innovation and implementation
The unique mode of technological innovation and exquisite design thinking are the core of the continuous growth of open computing; and the construction of standards is the basis for the implementation of open computing. It is precisely because open computing pays equal attention to the development of technology and standards that more and more users can enjoy the dividends of open technology, and the spirit of open source and openness is more deeply rooted in the hearts of the people.
This summit showcased a number of popular data center technologies including CXL high-speed switching technology, SONiC open network, 54V DC power supply, titanium power supply, liquid cooling, etc. These technologies are giving more scene innovation to open data centers possible. At the same time, the summit also released the "Open Acceleration Specification AI Server Design Guide" for generative AI application scenarios, further developing and improving the design theory and design method of the Open Acceleration Specification AI server, and speeding up the process of multi-computing inclusiveness.
Memory decoupling and pooling has always been a hot and difficult point in the industry, but the emergence of the serial cache coherence bus represented by CXL provides a low-latency access path and cache coherence between the host and remote shared memory Guaranteed, it provides the possibility for large-scale memory expansion and memory resource pooling. Wei Dong, director of the CXL Alliance, said that CXL technology is a new type of high-speed switching technology designed to provide higher data throughput and lower latency to meet the needs of modern computing and storage systems. The goal of CXL is to eliminate the memory gap between the CPU and the device, and between the device and the device. Through the decoupling of the memory and the host, logically build a memory resource pool that can be shared remotely, allowing multiple hosts to access the same memory pool, greatly Improve the efficiency of data exchange and realize dynamic memory expansion.
It is reported that Samsung demonstrated 512GB CXL DRAM that supports CXL 2.0, supports PCIe 5.0 interface, adopts EDSFF (E3.S), and accelerates the commercialization of next-generation storage solutions, thereby helping enterprises reinvest limited resources in enhancing Go to the server memory to accelerate the expansion of the CXL ecosystem. The converged architecture 3.0 prototype system released by Inspur Information at this summit and displayed on site has realized the complete decoupling and pooling of core IT resources such as computing resources, storage resources, memory resources, and heterogeneous acceleration resources with a groundbreaking system architecture design. Hardware reconfiguration has subverted the traditional architecture of the data center. Among them, based on CXL high-speed switching technology, the converged architecture 3.0 prototype system is the first to realize advanced features such as memory resource pooling, sub-microsecond level remote memory access, and fine-grained multi-host sharing in a single system.
As more and more services are digitized, the network traffic in the data center has surged, resulting in greater demand for network bandwidth. Network decoupling has thus become a realistic and urgent need for flexible expansion of network resources and agile O&M. . The open network realizes the decoupling of network software and hardware through the separation of network hardware devices and software codes, creating a more flexible, agile and programmable network architecture. SONiC initiated by OCP has become a mainstream open network technology and has been supported by more and more enterprises and organizations. Currently, there are more than 60 community members and more than 100 supported platforms. In the future, in addition to continuing to enhance programmability and support for end computing, SONiC will focus on enhancing support for network intelligent hardware.
Generative AI is one of the most concerned technical fields at the moment. At the summit, the "Open Acceleration Specification AI Server Design Guide" was released for generative AI application scenarios to help community members efficiently develop AI accelerator cards that comply with the Open Acceleration Specification, and greatly Shorten the adaptation cycle with the AI server, and provide users with AI computing power product solutions that best match the application scenarios. The Open Acceleration Specification (OAM) is promoted by the OAI group under the OCP. This standard is mainly used to guide the design of AI hardware acceleration modules and systems.
In addition to OAM standards, OCP standard specifications have covered all aspects of data center infrastructure, including computer rooms, cabinets, servers, storage, network equipment, power supplies, liquid cooling, cloud hardware, and more. The cooling environment project established in 2022 has become the largest cross-industry cooperation project of OCP so far. Representatives of many companies and even multiple industries have jointly focused on the innovation of data center liquid cooling technology. The project integrates 5 subprojects including Advanced Cooling Solutions (ACS) and Advanced Cooling Facility (ACF), such as ACS Cold Plate Subproject, ACS Portal Heat Exchanger Subproject, ACS Submerged Liquid Cooling Subproject, Waste Heat Utilizing sub-projects and more, the goal is to achieve standardization between the above-mentioned sub-projects and physical interfaces through cross-project coordination between the different cooling methods of the data center, thereby accelerating the pace of innovation in, near and around the data center.
In fact, more and more OCP community members and users are paying more and more attention to technical solutions for sustainable development of data centers, and sustainable development has become a community consensus. Zhao Shuai, General Manager of Inspur Information Server Product Line, believes that promoting the sustainable development of data centers has become a necessary action, from the development of environmentally friendly products, to the use of clean energy to improve the energy structure, to the application of liquid cooling technology to reduce carbon emissions, and even the recycling of electronic waste and reuse, the concept of sustainable development should be implemented throughout the life cycle of the data center.
Driven by the idea of collaborative development of software and hardware, the integration of open source software and open hardware is accelerating, and open source technology will profoundly change the division of labor and cooperation mode of the global IT industry. In the future, more and more cross-community cooperation will further break the boundaries of innovation, and endow infinite possibilities for business innovation with a more open data center.