Shulou(Shulou.com)11/24 Report--

Hello, everyone. I'm Mr. Jujube. In previous articles, I gave you a detailed introduction to the nature of computing and computing networks. Today is the last installment of the computing power series. Let's take a closer look at the networking architecture and technical system of the computing power network.

█ computing network architecture I mentioned before, the role of computing network is to provide users with computing resources services. It will thoroughly "integrate" computing resources into the communication network, and provide computing resources services that best meet the needs of users in a more integrated form.

This is a highly flexible and intelligent network.

The core feature of the computing power network is that it realizes the full takeover of the computing power resources and network resources through the computing power, and allows the network to perceive the users' computing power needs and its own computing power state in real time. After analysis, the computing power network can dispatch different locations and different types of computing resources to serve users.

We can think of it as a "universal faucet". It can always provide the right intensity, the right temperature, the right pH of the water flow according to your needs, so that you can have the best water experience.

In that case, we will naturally be very curious, what kind of "water net" is behind this "universal faucet"? How does it control and schedule resources?

Next, let's take a look at the architecture of the computing network.

First of all, I need to make it clear in advance that the computing network is still a technology under study and has not been completely finalized. Therefore, what I give below is only the computing network architecture (assumption) diagram of various operators and equipment vendors, and there will be some differences between them. Moreover, changes are likely to take place in the future, please know.

The following are representative diagrams of the computing network architecture of operators and equipment vendors:

China Mobile (from the White Paper on Computing Network Technology, 2022)

China Mobile (from the "Computing Network White Paper")

China Unicom (from Huawei WIN-WIN Innovation week, 2022)

China Unicom Computing Network Architecture (from the White Paper on Computing Network Architecture and Technology system)

Target Technical Architecture of Cloud Network Convergence of China Telecom

(from the "Cloud Convergence 2030 Technology White Paper")

Huawei (from Huawei WIN-WIN Innovation week, 2022)

These pictures are a little complicated, I believe everyone can not quite understand. I'll draw a simple schematic diagram, which you should be able to understand:

To put it simply, it is still the hierarchical idea of information and communication. according to the level, we can sort out the train of thought. In order to reduce the pressure of reading, you can first pass the green low-carbon and safety protection on the right.

Let's look at it from the bottom up, floor by floor.

Computing network infrastructure

The lowest layer of the computing network is still the infrastructure layer of the computing network, which cannot be escaped. In the computing network, infrastructure is no longer just computing resources (CPU, storage, GPU, etc.), but also includes network resources.

In terms of computing resources, the integration of computing infrastructure is still based on cloud technology, including cloud native, virtualization, containers and so on.

Here, some low-level new technological innovations related to computing power are also worthy of attention, such as heterogeneous computing, serverless computing (Serverless Computing), integrated deposit and computing, computing unloading and so on. You can ignore it for a while and give a presentation later.

On this side of the network, it has changed a lot. Previous articles have introduced you to the network changes under the trend of cloud-network convergence, which can be explained in one sentence, that is, to make a closed network more open (decoupled), making it like a computer, through the way of software (software definition). To give it a "command".

The bottom layer of the network (physical layer, data link layer), on this side of the transmission network, is still light (only light can ensure basic hard indicators such as communication bandwidth and capacity). The underlying optical technologies, such as OTN, OXC, ROADM, Clipl band, 400G / 800G and so on, are still them, there is no difference.

On the other side of the access network, 5G NR, PON (50G-PON), there is nothing special (except that the edge computing power will have something to do with it).

The ones that have changed greatly are at the top of the net. We'll talk about it later.

Integration of computing and network (computing resource management, network resource management)

Further up, it's starting to be different. Because the computing network needs to dispatch computing power and network resources, it is necessary to "preprocess" and "link up" the resources. (this layer, in fact, many structures are attributed to the orchestration management. I think it will be easier to understand if you bring it out alone. )

On this layer, the most noteworthy is computing routing.

The biggest difference between the computing network and the traditional network lies in its three characteristics. The first major feature is computing routing.

Before calculating power, we should first talk about the perception of arithmetic power.

Computing power awareness is the network's comprehensive perception of the deployment location, real-time status, load information and business requirements of computing resources and computing services. China Mobile's white paper on computing power perception network is analyzed from the perspective of computing power perception.

The Architecture of Computing-aware Network

(from China Mobile's "White Paper on Computational Awareness Network CAN Technology 2021")

On the one hand, each computing node publishes the computing network information metrics after modeling, and the network aggregates the computing network information reported by multiple nodes to build a global unified state view of the computing network. On the other hand, the network completes the unified analysis of the requirements of the service computing network, realizes the comprehensive perception of the service, and provides a guarantee for computing scheduling based on business requirements.

Power state awareness and announcement

(from the "White Paper on Power-aware Network CAN Technology 2021")

To put it bluntly, the purpose of power perception is to find out how many resources are on hand and what the needs of users are.

Perceived objects definitely need to be measured and identified. As a result, there is a calculation force measurement and calculation power identification. At present, operators attach great importance to these two pieces.

There is diversity in computing power, as well as heterogeneous hardware chips, so a unified measurement system and standards have not been formed yet. This will cause trouble for the establishment of a unified computing model. Very simple, the unit of measurement is different, how to carry out power evaluation and billing?

If you mark the power, it is equivalent to the ID card of the power. The user indicates the required service through the arithmetic power identification, and the network obtains the target computing power service, computing power demand and other information by parsing the computing power identification, which provides the basis for computing power scheduling. This numeric identification must be globally unified and verifiable.

All right, next, it's power routing.

In the computing-aware network architecture diagram of China Mobile just now, in addition to the computing resource layer, there is also a separate computing routing layer.

The computing routing layer is the core function layer of the computing aware network, which supports the perception and advertisement of network, computing, storage and other multi-dimensional resources and services, and realizes the joint scheduling of "network + computing".

The computing routing layer, including computing routing control technology and computing routing forwarding technology, realizes the on-demand scheduling of business requests in the routing layer.

When it comes to computing routing, we must talk about CFN (Computing First Networking, Computing priority Network).

In a broad sense, CFN is sometimes used to refer to the whole computing network. However, here, CFN is a distributed routing protocol that schedules and balances computing power between different edge clouds on demand.

CFN Technology Architecture (from the "White Paper on Computing-aware Network Technology")

CFN is actually a very interesting thing. The routing that we traditionally understand is the routing of data packets. The routing here in CFN publishes the status of computing resources and network resources as routing information to the network. Then, based on the virtual service ID, the network routes the computing task message to the most appropriate computing node.

At present, CFN protocol mainly spreads the load of computing nodes to the whole network in real time by carrying computing information in the way of extended byte information in the BGP header of service routing.

Obviously, computing routing and CFN are the prerequisites for computing scheduling in the upper layer.

Arrangement management

Further up is the core and key part of the computing network-choreography management.

Students who have engaged in the core network NFV (Network element function Virtualization) must be very familiar with the arrangement. In the NFV architecture, there is choreography. This is a very interesting word that "includes and arranges" you. To put it bluntly, it means complete control and scheduling.

The second feature of computing network is computing scheduling.

Computing power scheduling is based on the computing network brain to carry out intelligent scheduling and flexible scheduling of computing resources in the whole network.

More specifically, it is to complete the operation of computing power and the arrangement of computing services, complete the management of computing resources and network resources, including the perception, measurement and OAM management of computing resources, and realize the operation of end-users' computing network, as well as the management of computing routing layer and network resource layer.

After computing resources are perceived, identified and perceived, in order to manage, there is also a logical process of computing resource management (also known separately as: computing management). It mainly includes computing power registration, computing OAM (performance monitoring, fault management), computing operation (billing management, service contract, etc.).

Computing power management

(from the "White Paper on Power-aware Network CAN Technology 2021")

This function division is very troublesome. In some operator architecture diagrams, this piece is defined as the computing management layer, which exists alone. In some architecture diagrams, there is no such layer, which is delineated in the orchestration management. Therefore, each understanding is different, and it is very big. Anyway, you should know that there is such a functional group.

If you want to achieve it, you can't do without a "nerve center" that controls the network.

Scheduling and scheduling the core control part of the computing network, some operators call it the "computing network brain".

This "computing network brain" has four important functions:

1. Obtain the global real-time computing, network and digital resources, as well as the distribution of clouds, edges and ends, and build a global situation awareness map.

2. Cross-domain cooperative scheduling, which intelligently and automatically decomposes the scheduling tasks of multi-domain cooperation into each enabling platform to realize the resource scheduling of calculation, network and number.

3. Multi-domain fusion choreography, according to the needs of multi-domain fusion business, the atomic capabilities based on calculation, network and number are flexibly combined and arranged on demand.

4. Intelligent assistant decision-making, based on the SLA requirements of different services, the overall network load, the distribution of available computing resources and other factors, intelligently and dynamically calculate the optimal cooperative strategy of calculation, network and number.

The computing network manages the scheduling layer, maintains the global static computing power, service and network topology information, and synchronizes to each ingress computing routing node. The computing routing node maintains the topology information of the computing service and the real-time status information of the computing resources and the network resources, and carries out the collaborative scheduling of the computing network through the distributed computing routing nodes.

The "computing network brain" is the nerve center of the computing network, and the importance of co-ordinating the overall resources is self-evident.

It must be needless to say that AI artificial intelligence is introduced into the "computing network brain". In addition to AI, now, operators and equipment manufacturers also plan to plug "digital twins" and "intention engines" into them to try their best to improve the intelligent level of the "computing network brain".

To tell you the truth, if you look at the management orchestration layer carefully, I have a feeling that the computing network is the upgrade and integration of SDN and NFV technology, and it is the shadow of these two technologies everywhere. However, many other technologies are added to the surface to form a more complete and larger system.

Service operation

Is it gone on top of the calculation brain? No, don't forget, we have a service operator.

The service operation layer, to put it bluntly, is to provide the upper application with an interface to manage and control the computing network. Computer network brain is not a real boss, just a "big housekeeper". It is the upper users and applications that really control the network.

The purpose of the service operation layer is to open up the ability to provide computing networks. Key technologies include: computing power trading, computing power integration, computing power encapsulation, intention perception and so on.

The third feature of the computing network is the computing transaction.

Routing, scheduling and transaction are the three characteristics of computing network.

Computing transaction is an important innovation of computing network.

Before the convergence of computing power and the network, users will give money to whoever uses cloud computing and who uses the communications network. After the integration of the computing network, the computing power is everywhere, dispatching around, who should you give money to?

The purpose of arithmetic trading is to solve this problem.

In the computing trading scenario, there will be a credible ubiquitous computing trading platform.

From "Computing Network-Network Architecture and key Technologies in the era of Cloud Convergence 2.0"

The process goes something like this:

First of all, the consumer of arithmetic puts forward the detailed demand of arithmetic. Then, according to the demand, the computing trading platform integrates the available resources to generate a computing network resource table. According to factors such as delay and cost, there will be a corresponding quotation. Next, according to the resource table, the math consumer chooses a "package" that is most suitable for him. Then, the trading contract is signed on the arithmetic trading platform. Then, according to the contract, the trading platform schedules resources through the computing network, and updates the resource information. Finally, the contract ends, the service terminates, and resources are released.

Computing power network trading platform

(from "integrating multi-party computing networks to achieve optimal utilization of computing resources", Lei Bo)

In the 21st century, whenever trading mode is mentioned, it is bound to involve a word-blockchain.

Yes, that's right, the computing trading platform, also with the help of blockchain technology, to achieve distributed account books, anonymous transactions and other functions.

To sum up, the computing power network closely revolves around the computing power needs of users, covering the whole life cycle of computing power generation, scheduling, transaction and consumption. What the computing network wants to achieve is to move with people and "demand" to achieve the perfect match between supply and demand of computing power, which is neither easy nor wasteful at all.

The development of the standard of █ computing network in recent years, the computing network is very popular in China. Both operators and equipment manufacturers attach great importance to this concept.

Especially the domestic operators, the computing network is simply to the point of enthusiasm.

The reason is very simple, foreign operators seldom touch the network, basically engage in communications, at most engage in the cloud of traditional communications. On the other hand, domestic operators, which have both networks and clouds, particularly hope to integrate and activate the resources of the network and the cloud. Moreover, our country pays more attention to infrastructure construction, and computing network is an important digital infrastructure, so operators, as state-owned enterprises, also pay special attention to the accumulation of this area.

On the one hand, in order to grasp the initiative of technical standards, on the other hand, in order to promote the brand and image, the three major operators began to roll up on the computing network.

2019 is recognized as the first year of the computer network. On November 1 of this year, China Unicom took the lead in issuing the "China Unicom Computing Network White Paper" during the PT exhibition in Beijing. This is the first white paper related to the computing network in China.

Since then, domestic operators have scrambled to issue a number of white papers and actively promote the development of standards for computing networks.

In terms of standard promotion, I have compiled a table based on the information on the Internet, which roughly records the situation of the major standards organizations in the development of computing network standards, for reference only:

Comprehensive from CITIC Securities, major white papers, official website news

Progress of the experiment:

From Citic Securities

█ conclusion from the actual situation, the computing network is still a long way away from us. At present, we are in the development stage of cloud-network integration, and the integration of computing network or computing network is still under research.

In spite of this, as the future development trend of computing power and linkage, there is no doubt about it. The integration of calculation and network is inevitable, and it is also the return road of the whole communication industry. Traditional communication as a "pipeline", can not enhance their own value, only fully into the embrace of the cloud, is the best way out.

All right, that's all for today's article. There is a lot of technical information, which seems to be more laborious. Some things are my own understanding and may not be accurate. You are welcome to criticize and correct them.

Extended reading:

With regard to "arithmetic", this article is worth reading

What on earth is a "computing network"?

What exactly is "cloud-network convergence"?

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.