Key Element of the Computing Network – Networking
Deputy Director and Chief Scientist of China Unicom Research Institute, Tang Xiongyan:
Implementing integrated computing and networking services faces various challenges
The goal of the “Eastern Data and Western Computing” project is significantly different from the past simple data remote backup. “Eastern Data and Western Computing” aims to transfer data from the east to the west for computation and analysis, not just for storage and backup. Therefore, it is necessary to balance computing power and connect computing resources across time and space dimensions, allowing the west to leverage its cost and scale advantages in computing resources, fully activating and releasing computing capabilities. This requires the collaborative action of multiple elements such as networks, data centers, computing platforms, applications, and operational models.
Implementing integrated computing and networking services also faces multiple challenges. First, the business scenarios for “Eastern Data and Western Computing” are not yet clear; it is still necessary for all parties to explore which business applications are suitable for this initiative. The commercial value of “Eastern Data and Western Computing” still needs further exploration, and the utilization rate of computing resources in the west urgently needs improvement. Second, the technology and service standards for computing networks are not yet complete. The computing and networking industries are two significantly different sectors; both technology standards and industrial advancement require active collaboration between the two. Technical issues such as computing power measurement and transactions, network perception and routing, and joint scheduling and orchestration of computing and networking still require industry-wide collaboration to tackle. Third, the multi-party open sharing of computing resources still requires exploration. How to ensure the security of shared computing services, how to encourage participation through computing transactions, and how to provide differentiated network guarantees for different computing providers still involve many technical and commercial issues that need resolution.
Deputy Director of China Mobile Research Institute, Duan Xiaodong:
The sixteen-character guideline to achieve computing network objectives
China Mobile proposed a sixteen-character goal for the computing network: “Ubiquitous Computing, Symbiotic Computing and Networking, Intelligent Orchestration, Integrated Services.” The realization of ubiquitous computing involves three interconnected processes: first, physical integration, achieving physical connectivity between eastern and western computing; second, logical integration, connecting edge and center to achieve logical integration of cloud, edge, and terminal; third, heterogeneous integration, achieving heterogeneous integration through general computing. Through these three integrations, the goal of greatly ubiquitous computing can ultimately be achieved. Symbiotic computing and networking aims to gradually promote the integration of computing and networking into a unified entity. Computing and networking orchestration seeks to create an intelligent computing and networking brain, achieving overall network optimization through unified orchestration of cloud, data, intelligence, and manufacturing. Integrated services aim to provide services through the fusion of multiple elements, through the networking of social computing resources, and through the integration of data and intelligent services.
Director of Century Internet Research Institute, Ma Ju:
How to apply a new type of computing network based on blockchain
The first application scenario is personal data centers. The personal data center is a concept originally proposed by Century Internet, centered on returning data to the people, with blockchain technology as the security guarantee, and based on distributed storage and computing, promoting the confirmation, circulation, and secondary development of personal data identity and big data assets, a new type of information infrastructure.
Secondly, it can be applied in the construction of state-owned cloud in a hybrid cloud model. Through blockchain technology, distributed data centers, high-speed data centers, direct-link networks, standardized cloud-native platforms, and specialized unified operation and maintenance platforms can be integrated into the overall architecture of state-owned cloud, forming a joint computing platform that integrates multiple resources.
Finally, it can be applied to self-controllable data security. In an industrial environment, the ultra-connected new computing network can assist in ensuring the self-control and autonomous advancement of data throughout the entire lifecycle of design, production, and manufacturing, thus achieving sustainability and upgradeability. Additionally, new vehicle networking could also be an application scenario.
Key Element of the Computing Network – Storage
Senior Product Director of Alibaba Cloud Basic Products, Chen Qikun:
The storage industry is advancing amidst opportunities and challenges
The storage industry chain is mainly divided into hardware and software. In terms of storage software, many achievements have reached the forefront of the world. For example, Alibaba Cloud’s self-developed large-scale distributed storage system “Pangu” has transformed general storage server hardware into a large-scale, stable, and reliable storage system, with a single cluster scale exceeding 100,000 units, recognized as a leading technological achievement in the global internet. ESSD is the world’s first block storage product to enter the million IOPS era, with a single access latency as low as 30 microseconds. In terms of storage hardware, it involves many underlying industry chains, such as all-flash chip design, chip manufacturing processes, and disk technology, which are exploring forward amidst challenges.
As long as data exists, the prospects for the storage industry are very promising. For instance, the future of AI relies on massive data support, which will bring huge opportunities to the storage industry. Currently, the biggest challenge facing the storage industry is the network; only with more inclusive and efficient transmission can data flow at higher speeds and lower costs.
Senior Data Security Expert at Shuduo Technology, Pei Chao:
“Storage” and “Computation” are always mutually supportive and developed in coordination
Storage technology will not singularly move towards either “integrated storage and computation” or “separate storage and computation”. The core difference lies in which is central in the application scenario, “storage” or “computation”. For storage-centric cases like data centers, separate storage and computation remain suitable and relevant, but for computation-centric cases like data science and artificial intelligence, designing for integrated storage and computation can greatly reduce the losses caused by data transfer operations. In the future, regardless of whether storage technology develops towards “separate storage and computation” or “integrated storage and computation”, it will need to address different scenarios, providing different applications and services based on the structural characteristics between “storage” and “computation”, as well as their dominant positions.
Key Element of the Computing Network – Computation
Director of Intel China Research Institute, Song Jiqiang:
Establishing a more complete heterogeneous ecosystem is crucial
Heterogeneity has become a new competitive point globally. Now mainstream chip suppliers are seeking to make their heterogeneous layouts more complete. To stand out in heterogeneous competition: first, there should be a variety of architecture accumulations; second, different resources should be available for use at different process nodes; third, early layout and planning of packaging-level technologies are needed. Heterogeneity requires layout architecture, production processes, and packaging technologies, all aimed at establishing a more complete heterogeneous ecosystem in the future. This ecosystem will include different architecture IP modules or chiplet modules, capable of producing these modules at the best cost-performance ratio across various process technologies, and connecting them using the most standardized methods; fourth, a convenient and user-friendly software system is needed, where upper-level users can specify functional requirements, and the lower levels can automatically migrate with heterogeneous changes.
Founder and CEO of Suiyuan Technology, Zhao Lidong:
Heterogeneous computing is a key support and foundation for data-centric intelligent industries
Heterogeneous computing plays a key supporting and foundational role in data-centric intelligent industries. Given the high barriers, advanced technology, and long cycles of the chip industry, it will become a key technology and a crucial part of national technological competitiveness. The competitive direction is to improve the industrial chain through key technology and product innovation under the guidance of relevant industry norms and standards, as well as to leverage the corresponding industrial agglomeration and scale effects.
Currently, China’s integrated circuit industry is in a stage of vigorous development. Whether from the horizontal variety of heterogeneous computing chips or the vertical tracks of manufacturers, there is a flourishing diversity, and each manufacturer’s underlying hardware architecture and programming models have their unique strengths, requiring more guidance from the standards and norms level to steer the market direction.
Vice President of TianShu Intelligent Chip, Guo Wei:
The domestic general computing industry ecosystem based on self-developed chips needs improvement
In “Eastern Data and Western Computing”, “data” refers to data, and “computation” refers to computing power, that is, the ability to process data. Data centers can be divided into three core elements: computing, storage, and networking. Therefore, the chips required for these three core functions have become the focus of attention in the industry under the “Eastern Data and Western Computing” strategy, including general computing products such as general GPUs, which will provide a strong boost for infrastructure development.
Our country still has two areas to improve in general computing: first, there are significant risks in the industrial chain and supply chain. The general GPU market is dominated by NVIDIA and AMD, chip design tools rely on foreign companies, wafer foundries cannot be self-sufficient, and key components such as packaging substrates depend on imports, leading to risks in the general computing chip industrial chain and supply chain.
Second, the industrial ecosystem has not been effectively constructed. The domestic general computing chip sector started relatively late, with most innovative companies established only in recent years, small in scale and limited in strength, facing difficulties and long cycles in developing software stacks, algorithms, and toolchains. Most of the computing chip products that have been released are still in the stages of adaptation verification and bulk shipment, awaiting large-scale market validation, and an industrial ecosystem based on self-developed chips has yet to form.
Key Opportunity of the Computing Network – “Eastern Data and Western Computing”
Director of Planning Department, National Information Center Big Data Development Department, Wang Jiandong:
Strengthening the top-level design of the national integrated big data center requires five “data”
“Data Network”, which means optimizing the construction layout of data centers, focusing on solving the current serious issues of resource supply and demand imbalance in regional digital economy infrastructure, and enhancing the centralized and coordinated construction level of data centers and networks in eastern and western regions, effectively promoting “Eastern Data and Western Computing”.
“Data Link”, which means promoting the service-oriented nature of computing resources, focusing on solving issues such as inconsistent application standards and interfaces among public cloud solution providers like Huawei and Alibaba, by constructing a universal, integrated cloud service interface similar to fourth-party logistics, providing enterprises with more convenient, cost-effective, and efficient cloud services.
“Data Chain”, which means accelerating the cultivation of the data element market, focusing on solving the lack of credible circulation requirements for data between government-government, government-enterprise, and enterprise-enterprise, constructing mechanisms for data quality assessment, credible circulation, and joint modeling, achieving the resource, asset, and capital levels of data element circulation and distribution mechanisms, supporting the construction of a unified large market for data elements.
“Data Brain”, which means deepening collaborative application innovation of big data, focusing on addressing the insufficient perception capability of government and enterprises regarding economic and social operations, and the low level of macro decision-making and risk assessment, cultivating a data-driven, collaborative, integrated, and shared big data application ecosystem, and building a data decision-making brain.
“Data Shield”, which means strengthening big data security protection capabilities, focusing on addressing a series of challenges and issues emerging in the field of data security, by establishing an integrated security protection system for networks and data and a comprehensive regulatory system for blockchain, effectively improving the overall security level of digital economic development.
Chairman and President of Huayun Data, Xu Guangbin:
The essential goal of “Eastern Data and Western Computing” is to achieve a dual transfer of computing resources
Computing power is the core productive force of the digital economy and has become a new focus of global strategic competition. As digital technology continues to penetrate all fields of the economy and society, the overall demand for computing power is expected to continue to grow rapidly by over 20% annually. The essential goal of “Eastern Data and Western Computing” is to achieve a dual transfer of computing resources. In simple terms, “Eastern Data and Western Computing” aims to allow the computing resources in the west to better support the computation of eastern data, thereby empowering digital development. Therefore, “Eastern Data and Western Computing” seeks to rationally optimize the resource allocation in the central and western regions, deploying computing power in a layered, orderly, and staggered manner, rather than simply focusing on the development of the west. The Yangtze River Delta economic area has a large volume and many high-end businesses, such as financial securities, artificial intelligence, remote medical care, and industrial internet, which require high network standards, making it more aligned with realistic demands and development needs to layout hub nodes in the Yangtze River Delta.
Technical Director of UCloud Infrastructure Center, Fu Dongming:
Four key points to focus on for the full implementation of “Eastern Data and Western Computing”
First, strengthen the connectivity of network facilities. Accelerate the establishment of direct data connection channels between the east and west, creating a number of demonstration lines for “Eastern Data and Western Computing”. Optimize the communication network structure, enhance the network node level of national data center clusters, and improve network transmission quality.
Second, strengthen the linkage of energy layout. Enhance the integrated design of data centers and power grids, promoting renewable energy companies to supply power to data centers. Support data center clusters with accompanying renewable energy power stations. Prioritize energy consumption indicators for data center projects that achieve significant results in implementing “Eastern Data and Western Computing”.
Third, support the integration of technological innovation. Encourage innovations and models in data center energy conservation and carbon reduction, renewable energy power supply, heterogeneous computing integration, cloud-network integration, multi-cloud scheduling, and data security circulation, enhancing support for the research and development of key technological products and their large-scale applications.
Fourth, promote the growth of the industrial ecosystem. Support the improvement of the data center industry ecosystem, strengthening the upstream equipment manufacturing industry and downstream data element circulation, innovative applications, and new consumption industries around data centers. Support the western computing hubs to develop labor-intensive industries such as data processing, data cleaning, and data content services around data centers.