What exactly is the “computing power network”?

What is a “computing network”?

Let’s get straight to the point. What exactly is a computing power network?

The computing network is not a specific technology, nor is it a specific device. From a macro perspective, it is an idea, a concept. From a micro perspective, it is still a network, a network with a completely different architecture and nature.

The core purpose of the computing network is to provide computing resource services to users. However, its implementation method is different from the traditional method of "cloud computing + communication network". Instead, it completely "integrates" computing resources into the communication network to provide computing resource services that best meet user needs in a more holistic form. ​

Therefore, some people call the computing power network "Network As A Computer".

What is in front of you is like a computing machine. You don’t need to care what is behind it, you just need to know that it can provide you with the computing resources that best meet your needs.

The Evolution of the Relationship between Computing and Networking

Just reading the previous paragraph may make you feel dizzy. Next, let’s start from the origin of the Internet.

In the article on computing power, I told you about the development history of computing power, which is actually the development history of computers.

The development of the Internet is actually closely related to computers. (The Internet mentioned here refers to the data communication network, not the voice telephone network. The same applies below.)

In the 1960s, ARPANET (also known as the famous "ARPANET", the predecessor of the Internet) was established to enable large computers in major American universities to transmit data.

The ARPANET connecting four universities in the Midwestern United States (1969)

In the 1970s, early local area network technology (Ethernet, TCP/IP protocol) was developed to serve computer-to-computer communications in universities and research institutions.

The emergence of the Internet, in addition to enabling peer-to-peer (user-to-user) information exchange, is more important in that it makes some complex and high-end computing capabilities accessible to ordinary users.

In that era, the computing power of users' PCs was very weak, the things they could do were very limited, and there were few content resources (hard disk storage capacity was very small).

With the Internet, users can establish a connection with the computer room (data center), access the servers in the computer room, and share the server's CPU and storage.

Concentration and sharing of computing power

For complex and difficult computing tasks, they can also be assigned to different computers through the network to complete the computing tasks together. This is grid computing, a form of distributed computing.

After the 1980s, the number of networks increased and their scale became larger. As a result, people built backbone networks connecting major regions, and eventually formed the global Internet. ​

Small network becomes big network, that is the Internet

With the global Internet, the computer room that carries computing resources can become larger and more powerful, providing computing services to more users. This computer room has become an Internet Data Center (IDC).

After entering the 21st century, companies such as Amazon and Google took the lead in developing cloud computing based on Internet data centers in order to better manage massive servers (and also to achieve high-performance and high-reliability computing tasks with cheap servers).

The core of cloud computing is virtualization technology. To put it simply, all computing resources such as CPU, memory, hard disk, graphics card, etc. are turned into a "resource pool" and flexibly allocated to users.

Virtualization technology breaks down physical resources and turns them into virtual resources

On the Internet, huge changes are also happening simultaneously.

Initially, the focus of the network was on improving transmission rate, capacity, and coverage. After all, with more users, more data centers, and more Internet companies, the demand for bandwidth has increased. If you want users to access faster and have a better experience, you have to make the "water pipe" thicker.

During this period, optical communication and mobile communication have developed rapidly. The use of optical fiber can significantly expand the communication bandwidth. The use of mobile communication can achieve communication access anytime and anywhere.

By around 2010, our communication networks had basically achieved physical connections between people and between people and data centers.

It is not easy for Zaojun to draw pictures. Please indicate the source when reprinting.

At this time, with the emergence and maturity of cloud computing and big data technologies, the core tasks of communication technology began to change - the objects of communication connection began to expand from people to things, and the Internet began to expand from the consumer field to the industry field ( industrial manufacturing, transportation and logistics, banking and finance, education and medical care, etc. ).

Industrial Internet began to rise, and the Internet of Things also began to rise, thus opening the door to the digital transformation of human society as a whole.

Phase 1: Cloud-Network Collaboration

In the digital age, everything revolves around data.

​Experts have decisively defined all informationization, digitization, and networking as: "mining data value", "creating digital wealth", and "developing the digital economy".

IT technologies represented by cloud computing, big data, and artificial intelligence have been renamed computing power. Communication technologies represented by communications technologies have been renamed connectivity. They have become the most important tools for digital transformation. (Storage resources are also called "storage power," but they are generally classified as computing power.)

The process of mining data value

In this era, all computer software and hardware have been abstracted and turned into resources like water and electricity, called "computing power resources."

All applications, such as watching TV shows, playing games, office automation, AR/VR, etc., are also collectively referred to as using "computing power applications" and enjoying "computing power services."

Computing power has become an important productivity that the entire society needs.

However, computing power is very different from electricity. Electricity is energy, and you can use it as long as the power grid is connected. But computing power has different attributes and types. Different users and different scenarios have different demands for computing power.

In other words, there is diversity in computing power.

The article on computing power mentioned that there are different types of computing power, including general computing power, supercomputing power, and intelligent computing power.

For example, if I play the game PUBG, I need game algorithms and graphics rendering. But what you provide is intelligent computing power. Is that appropriate?

For another example, I am working on the Internet of Things for street lights, controlling the on and off of street lights, which is a very simple operation. As a result, you provide expensive supercomputing power and assign me Tianhe-1. Is this appropriate?

For another example, if I want to mine Bitcoin, and you provide me with x86 CPU general computing power, but the mining efficiency is extremely low, is this appropriate?

Obviously neither is appropriate.

Some people want high-performance computing power, some want computing power with fast response speed (low latency), and some want cheap computing power... Cloud computing alone cannot flexibly meet the diverse needs of users.

Therefore, when it comes to computing power, we thought of the cooperation of the network.

On the other hand, the network (communication operators) also have a strong desire to cooperate.

Here are the reasons:

1. Traditional networks are too closed, and equipment vendors control the technology. Once selected, it is difficult to replace. IT technology represented by cloud computing emphasizes open source, pooling, software, and decoupling of software and hardware. As the first party, operators can take more initiative.

2. Although traditional networks are in the field of communication, they also use computing power. In the field of routing and switching, and in the field of core networks, computing power is actually the main factor, and the equipment itself is an "x86-like server " . If you want to improve the operating efficiency of the equipment, you need to IT-based and software-based communication networks, which can simplify network operation and maintenance and achieve elastic scaling of capacity.

3. Operators are the main body of network operation, but after laying the network, they can only serve as a "pipeline" and cannot touch user data or user services. Services are more profitable than traffic packages, and operators do not want to be marginalized, so they hope to use the network as their capital to participate in the competition in the cloud market and share the cake.

4. Domestic operators have networks in their left hand and cloud in their right hand. However, operators’ clouds are relatively dependent on equipment vendors. By cooperating with cloud and network, operators can use cloud to sell networks, use network to sell cloud, and learn while selling, thus increasing their control over the cloud and not suffering any loss from either side.

Therefore, around 2010, Yunhe.com began to break down barriers and enter the first phase of cooperation. At this time, Yunhe.com was in its "first love" and both parties still emphasized their respective identities and cooperative relationships, so it was called the "cloud-network collaboration" stage.

The SDN (software defined network) and NFV (network function virtualization) that everyone is familiar with are typical representative technologies in the cloud-network collaboration stage.

At that time, SDN was mainly aimed at the bearer network. The management and forwarding functions of the bearer network routers were separated, and the management function was centralized. In this way, the network was software-based, and instructions could be issued at any time.

SDN, the network is disassembled

NFV is mainly aimed at the core network. It introduces cloud technology into the network, transforms the communication network unit from a professional device into a general-purpose x86 device, and implements network functions through virtual machines, making it more open and flexible.

NFV, migrating network element functions from physical devices to virtual devices (cloud services)

In fact, the wireless access network (base station) is also cloud-based. Antennas cannot be cloud-based (they still need to send and receive signals), but baseband computing can be cloud-based, so there are Open RAN, vRAN, C-RAN, etc. Due to space limitations, I will not introduce them in detail.

SDN and NFV are technologies and concepts that introduce cloud into communication networks, which is equivalent to using cloud to transform the network.

From the perspective of the cloud, we also get "benefits" from the network. This important "benefit" is MEC edge computing.

With the Internet, the cloud found that it could "flow" along the Internet. It sinks part of the computing power of the central cloud to various levels of the communication network, closer to the user, and can meet the user's demand for low-latency computing power.

This computing power can be in your home router, in the weak current machine room of the building, in the base station room, or in the computer rooms at district, county, and city levels. In short, it is everywhere.

Edge computing = computing power sinking

Edge computing has completely overturned the traditional computing architecture of either end or cloud, turning computing resources into a three-level model of "cloud, edge, and end". They work together to provide users with the required computing services.

As a result, the term "ubiquitous computing power" began to emerge.

In the era of cloud-network collaboration, the cloud can mobilize the network ("cloud-network"), and the network can also cooperate with the cloud. As mentioned above in SDN, the network is defined by software, and the network functions become options on the platform. When operating the cloud, you can call the network functions and configure the network by clicking a button.

Phase 2: Cloud-Network Integration

The emergence of cloud-network collaboration has revealed the direction of change for the entire ICT industry. Its initial achievements have also encouraged operators, equipment manufacturers, and cloud computing service providers.

A few years later, everyone agreed that it was not enough for the cloud and the Internet to simply work together, and that they should move towards comprehensive integration. In this way, "cloud-network integration" made its debut.

The fundamental reason for this change is actually the wave of digital transformation. As digitalization continues to deepen, data is becoming increasingly massive. In particular, the widespread implementation of data-centric artificial intelligence businesses has intensified the demand for computing power across society.

In order to meet the urgent demand for computing power, the integration of cloud and network must be accelerated.

At this stage, due to the emergence of edge computing, cloud computing can no longer represent computing power alone, so the word related to "cloud" has gradually become "computing". (The strong rise of intelligent computing and supercomputing has also made the word "computing" more powerful, more impressive, and more representative. )

The network, on the other hand, has completely lost its status as an equal to the computing power and has begun to accelerate its “integration” with the computing power. In fact, to be frank, it is being “integrated” by the computing power.

Integration is the current action, and the ultimate goal of integration is, of course, to fully integrate computing and the Internet. In other words, in the future, we will achieve "computing and Internet integration".

The integrated “computing network” is the “computing power network”.

Did you understand the whole process? The reason why the concepts on the Internet are so mixed is that the three major operators and equipment manufacturers such as Huawei like to give names and hype concepts, and they are unwilling to admit or unify each other. Therefore, there are many different names in the media. In fact, many nouns have the same meaning.

Operators are really trying too hard to come up with words, such as "1+2+3" and "ABCDE", which are even more confusing.

There are also several English names for computing power networks. For example:

• CPN (Computing Power Network )
• CFN ( Computing First Networking )
• CFN (Computing Force Networking)
• CAN (Computing- aware Networking )

The most commonly used one at present is CFN (First).

At the beginning of this article, Xiaozaojun said that the purpose of the computing power network is to provide users with the most suitable computing power resource services.

​This suitability refers to the matching of computing power types, appropriate computing power scale, and optimal computing power cost-effectiveness.

The core problem that the computing power network needs to solve is the problem of insufficient computing power supply across the entire network as the demand for computing power increases rapidly. ​

At present, Moore's Law has gradually reached a bottleneck, the space for improving the computing power of a single chip is getting narrower and narrower, and the cost is getting higher and higher. When the computing power of a single point cannot be continuously doubled, revitalizing the existing computing power resources is the only way to solve the problem of insufficient computing power.

Computing power can no longer keep up with the growth of data (Image from Yushu Technology)

In other words, making computing power flow, providing precise services to users, and improving computing power utilization are more valuable than simply piling up computing power and focusing on chip manufacturing processes.

The "East Data West Computing" that is very popular this year is a practical implementation of the computing power network concept.

The eastern region has a high demand for computing power, while the western region has a low computing power cost (low temperature, low cooling cost, and cheap energy). Therefore, with the help of a powerful communication network infrastructure, the computing power with low latency requirements can be migrated to the western region to achieve a more perfect computing power cost performance.

So, what kind of architecture is the computing power network? What underlying technologies is it based on? How are its three major features - computing power routing, computing power scheduling, and computing power trading - achieved? At present, what is the progress of the computing power network standard? ​