The computing power network has its own calculations

In recent years, the wave of digitalization has continued. Especially since last year, the rise of AI has led to a huge explosion in the industry's demand for computing power, and the construction of computing power infrastructure has become a key topic of concern in the industry.

On the one hand, the demand side has put forward more requirements for the supply of computing power, and GPU, an important component of intelligent computing power, has been stuck; on the other hand, the country's dual-carbon strategy requires a change in the traditional development model, emphasizing that computing power construction cannot be simple and crude. As a result, new names and concepts such as "Eastern Data and Western Computing" and "Computing Network" have appeared frequently, and huge business opportunities have emerged in the digital industry.

Amid the dazzling array of corporate releases and lengthy government documents, I discovered that the industry has not yet reached a consensus on how to build the computing infrastructure and the "computing network." Telecom operators, Internet companies, and the National Development and Reform Commission all have very different ideas.

【Telecom operators’ computing power network】

The first time I saw the name "computing power network" was at China Mobile.

When Yang Jie worked at China Telecom, he always insisted on promoting the "cloud-network integration" strategy and made Tianyi Cloud the most eye-catching cloud business among telecom operators.

After being transferred to the position of Chairman of China Mobile in 2019, his initial business direction was still centered on 5G, leveraging the strong advantages formed in the field of wireless communications to expand digital capabilities.

Since the second half of 2021, the development direction has gradually adjusted to a combination of "connection + computing power + capability", and "computing power network" has frequently appeared in China Mobile's documents.

According to reports, China Mobile first appeared in the public eye together with computing power at the ITU SG13 plenary meeting held in October 2019. At that time, the "Demand and Application Scenarios of Computing Power Aware Networks" project led by China Mobile was approved by the plenary meeting and became the first international standard project for computing power aware networks. The landmark event was the "China Mobile Computing Power Network White Paper" issued at the 2021 Partner Conference.

Later I discovered that as early as November 2019, China Unicom had released the "China Unicom Computing Power Network White Paper", but at that time the industry did not pay much attention to computing power, and the relatively weak China Unicom had a relatively small voice, which did not cause much response in the industry.

From a global perspective, as the ICT industry continues to evolve, telecom operators that were once at their peak in the 2G era have gradually lost their dominant position. After the rise of cloud computing, traditional telecom operators have been defeated in this field. The US operators that were once the industry benchmark have abandoned the public cloud business, and European ICT companies are even less able to compete with US cloud computing companies. Only China's three operators are still struggling to survive under the ridicule of the industry and the pressure of Internet companies.

In order to maintain the survival of cloud services, Chinese telecom operators generally adopt the "cloud + network" combination model, trying to use their own advantages in network to combine connection services with cloud services to form a digital foundation for enterprise customers. Such a combination of capabilities is dispensable for ToC digitalization and Internet-native companies; but it is still quite attractive to traditional industries, especially large-scale customers across multiple locations.

Nowadays, there is a strong call for the development of computing infrastructure, and telecom operators believe this is a new development opportunity:

Telecom operators have the ability to build infrastructure on a large scale, and they themselves have a large number of computer room resources. Through transformation and expansion, the infrastructure can not only provide connectivity and cloud computing capabilities, but also provide the computing power and storage capacity that customers need; combined with a nationwide transmission network, they can provide high-quality, reliable, and sufficient integrated computing and network services.

If we want to go a step further, we need to unify the management of computing network resources and intelligently orchestrate capabilities, continuously optimize the combination of connections and computing power, and improve the supply efficiency of digital infrastructure.

Therefore, telecom operators ambitiously believe that by building on their existing capabilities and resource advantages and using computing power networks as an important lever for their rise, they can regain dominance in the digital industry and create greater glories in the future.

Therefore, in the eyes of telecom operators, the computing network is a digital infrastructure that provides data and computing power for applications. Its essence is still the network. This network can convert connections into data, call computing power and storage power for customers, and lay the foundation for the development and operation of digital applications.

【Computing power network of Internet cloud companies】

Today, Internet companies are no longer as glorious as they were in previous years. ToC business has peaked, capital operations have been hindered, and technological and industrial innovations have encountered bottlenecks. In addition, the domestic and international situation has undergone major changes. Many factors have combined to make today's Internet companies much more low-key than before the epidemic.

Recently, the development of public cloud business of Internet companies has encountered certain setbacks: the development speed has declined, the customer scale has reached its peak, and some of the lost customers have been poached by other cloud platforms, and some have switched to private clouds. As the rapid expansion of business scale is unsustainable, funds are insufficient to support all-round expansion, marginalized expansion business is shrinking, and the dark cloud of cost reduction and efficiency increase hangs over the heads of employees. It is not easy to maintain the stable operation of a huge system, and risks are frequent.

How will the cloud business continue? If the competition is only at the IaaS level, the homogeneous competition will be very fierce. Facing the large and tough telecom operators, the original model of "subsidizing in exchange for market share, forming a monopoly and then making profits" is simply not feasible. Moreover, Internet companies focus on efficiency in their operations. Although they can concentrate their superior forces on key events, they are not advantageous in the long run when facing the telecom operators' marketing channels with wide coverage and operation models that go down to the grassroots level.

Therefore, when Internet companies engage in cloud business, they must rely on their technological advantages and the ability to maneuver flexibly in the market to engage in differentiated competition with giants.

Internet companies do not like unified standards, because in the IT and Internet circles, it is most beneficial to promote de facto standards based on their own interests. Whoever forms relatively solidified technical capabilities first, the latecomers can only adopt a follow-up strategy and comply with the standards set by the former if they want to cooperate and connect. This not only makes it easier for the leader to continue to lead, but also increases the cost of platform migration for partners: development done on platform A requires additional costs to migrate to platform B. If it is not necessary, who is willing to spend more money?

Developing business based on factual standards requires strong support from the industry ecosystem. If no one plays along and no one responds to the standards you make, then the money is wasted. Therefore, when a company is relatively weak and small, it is not suitable for this high-profile approach. A common phenomenon in the IT and Internet fields is that the strong will always be strong: the stronger you are, the more business opportunities you will have, the more active your partners will be, and the closer your ecosystem will be to success.

From a cost perspective, customers and application developers hope that cloud platforms provide capabilities that are standardized. If all cloud platforms provide the same services and capabilities, and the competition is highly homogeneous, customers and developers will take the initiative in the game with the platforms, forcing them into price competition.

The IT industry is highly open, and the advantages gained by technological leadership are difficult to maintain for a long time. New things that are carefully thought out may be quickly discovered and imitated by competitors. What benefits will the cloud have if it loses its premium ability?

Therefore, Internet companies can only start from the aspects of technology and services, continuously build the differentiated advantages of the platform at the PaaS layer, provide effective help and support for application developers and digital users, and guide partners to use as many platform capabilities as possible while using homogeneous infrastructure. The stronger the dependence on the differentiated capabilities provided by the platform, the lower the probability that partners will leave the network and switch to other platforms. By building an industrial ecosystem, cloud platforms can regain the dominant position in competition.

Today, the "computing power network" provides Internet companies with new opportunities to expand their cloud business: leveraging the industrial development opportunities of building digital infrastructure, realizing the distributed construction and networking of cloud resource pools, reducing costs through reasonable resource layout and efficient calling, and using their own technical capabilities to continuously optimize the operation and management of the entire network, and forming cloud computing capability outputs that can be flexibly adjusted according to user needs.

Therefore, for Internet cloud companies, computing power networks are the distributed cloud computing foundation that supports digital applications. It is an upgraded version of the cloud computing platform, which reduces the supply cost of cloud services through the reasonable allocation and flexible combination of capabilities and resources, thereby increasing the scale of cloud business.

[The computing power network in the mind of the National Development and Reform Commission]

The full name of the NDRC is "National Development and Reform Commission". It is an important government department responsible for formulating and issuing industrial policies. The understanding and conception of computing power networks are hidden in various documents and often require in-depth analysis to interpret.

In May 2021, the National Development and Reform Commission, together with the Cyberspace Administration of China, the Ministry of Industry and Information Technology, the National Energy Administration and other departments, issued the "Implementation Plan for the Computing Hub of the National Integrated Big Data Center Collaborative Innovation System", which mentioned the "East Data West Computing" project for the first time. At that time, the focus of this document was on the construction of the computing hub, and the content related to the network was only the "interconnection" in the key work.

In the more than two years since then, most relevant documents have focused on the construction of computing power hubs, until the end of last year when the "Implementation Opinions on Deeply Implementing the 'East Data West Computing' Project to Accelerate the Construction of a National Integrated Computing Power Network" was released, and the name "Computing Power Network" made its debut.

It is not difficult to see that the NDRC is now applying historical successful experiences and cross-industry practical cases to the field of digitalization. It has previously worked on the South-to-North Water Diversion Project and the West-to-East Power Transmission Project, and now it is familiar with the East-West Calculation Project.

Moreover, when introducing computing power, leaders and experts in the information industry often compare computing power to electricity, and rely on computing power infrastructure for public utilities, believing that future computing power output will be like hydropower, ready for immediate use. Therefore, from the perspective of the National Development and Reform Commission, it is logical to copy the dispatching and management model of electricity and hydropower to East Data West Computing and transplant the electricity reform plan to the construction of computing power infrastructure.

In the 2002 "Electricity System Reform Plan", the core of the reform is to establish an electricity market system, split the original State Power Corporation into two major power grid companies, five major power generation groups and four major auxiliary industry groups, and realize the separation of power generation, transmission and sales. So, can the development of computing power separate the supply and transmission of computing power? The former is the "computing power hub node" and the latter is the "computing power network". Then, "East Data West Computing" is used to balance the conflict between industrial development and energy conservation and emission reduction. These elements are combined to form a national integrated computing power infrastructure.

According to the idea in the NDRC document, computing power infrastructure should be decoupled.

The first level of decoupling is the decoupling of the computing power hub center and the computing power network. Referring to the power reform plan, some enterprises and units are responsible for the construction and operation of general computing power, intelligent computing power and other facilities; the other part of the enterprises is responsible for the construction and operation of the computing power network that realizes high bandwidth, low latency, and supports heterogeneous and cross-domain transmission.

This conflicts with the thinking of telecom operators.

The second layer of decoupling is to promote the standardization of computing power supply and reduce the migration costs of application developers and customers in the process of using computing power. Industry experts will take the lead in formulating computing power grid connection standards, promote the standardization and popularization of computing power services, and improve the usability of computing power resources; strengthen the interconnection and unified services of diversified computing power, promote the intelligent supply, scheduling, use and settlement of computing power, and improve the convenience of computing power use by small and medium-sized enterprises and the usability of users in using various types of computing power.

This is inconsistent with the thinking of Internet companies.

So if traditional digital infrastructure suppliers do not cooperate, will the plan fail? We saw the following in the NDRC document:

"Support the cultivation of professional computing network operators, strengthen the operation and management of computing power and networks in all aspects of operation, management and maintenance, explore a standard system of unified measurement, unified billing, unified transaction and unified settlement, and a computing network collaborative operation mechanism, and create a new computing service ecosystem with efficient supply, mechanism-based scheduling and intelligent operation."

“Explore the construction of a national integrated computing network prototype technology test field, strengthen the development of computing network technology standards, and support industry, academia, and research parties to carry out computing network common technology research and development and trial promotion. Explore new R&D models, gather high-end technical forces, establish a national computing network common technology R&D platform, and strengthen national strategic scientific and technological strength. Actively carry out the "unveiling and commanding" of key technologies such as distributed computing power parallel scheduling and heterogeneous computing power scheduling, and cultivate the computing power industry ecosystem. Give full play to the role of scientific research institutions, industry associations, and industry alliances, and promote the formulation and promotion of standards and specifications in the fields of computing power scheduling and computing power settlement.”

The industry pie is so big, even if you don't want to do it, there are plenty of people who are willing to do it. There will always be brave men when there is a big reward.

Therefore, for the National Development and Reform Commission, although the computing power network is an integrated infrastructure, its structure is loosely coupled. The ultimate goal is to achieve the goal of developing digitalization under the constraints of low-carbon and environmental protection, and to develop digitalization using inexpensive and high-quality computing power infrastructure.

Conclusion

The introduction of industrial policies, the proactive construction of large-scale infrastructure, and the promotion of industrial development with low-cost and high-quality infrastructure are important experiences in China's development of industrial economy. Large-scale high-speed rail construction has given wings to the economic development along the line, and the speed increase and fee reduction of telecom operators have promoted the explosion of the mobile Internet industry. These are all successful cases. Therefore, we have reason to believe that the construction of a national integrated computing network characterized by "East Data West Computing" will effectively reduce the computing power cost required for digitalization and create conditions for the overall development of the digital industry.

So, if East-West Computing fails to achieve the desired goal, the worst case scenario is that the infrastructure is not efficient enough and the computing cost cannot be significantly reduced. Even so, if greater value can be created on the application side, digitalization will still flourish, and a bigger pie will still bring great benefits to the participating players.

It’s just that at this stage, after seeing the roadmap clearly, outsiders can just sit back and wait for the results in a few years; but those involved must make a choice.