WO2022077197A1

WO2022077197A1 - Fusion processing device

Info

Publication number: WO2022077197A1
Application number: PCT/CN2020/120531
Authority: WO
Inventors: 韦超; 刘彩虹; 王浩; 周晓悦
Original assignee: 海能达通信股份有限公司
Priority date: 2020-10-13
Filing date: 2020-10-13
Publication date: 2022-04-21

Abstract

A fusion processing device. A back plate comprises a first location region, a second location region, and a third location region in which a network resource module, a general-purpose computation resource module, and a dedicated resource module can be respectively installed. Any two of the network resource module, the general-purpose computation resource module and the dedicated resource module are interconnected by means of a communication interface and a bus on the back plate. The network resource module, the general-purpose computing resource module and the dedicated resource module constitute hardware architecture for products such as a radio access network base station, a mobile edge computing device, and a miniaturized radio core network, such that different products can use the fusion processing device for hardware support, thereby unifying configurations of devices, and enabling maintenance to be performed easily. Moreover, modular constitution enables various modules to be configured and upgraded independently.

Description

Fusion processing equipment

technical field

The present application relates to the field of communication technologies, and in particular, to a fusion processing device.

Background technique

Existing radio access network base stations, mobile edge computing equipment and wireless miniaturized core network products all have their own device forms and are of various types.

For example, the wireless access network base stations of the current mainstream communication equipment manufacturers in the industry mainly use special equipment. The traditional wireless access network BBU baseband processing architecture is shown in Figure 1a and Figure 1b: The switching main control processing board (chip) and the multi-core processor chip (the multi-core processor chip is located in the clock switching main control processing board) are interconnected through a high-speed Ethernet interface to realize wireless access network baseband processing.

For another example, the distributed small cell DU base stations organized by the current mainstream 5G O-RAN in the industry are mainly based on general-purpose servers and FPGA/GPU/dedicated ASIC acceleration cards to realize business processing. For example, referring to Figures 1c and 1d, a COTS general server (motherboard), a baseband processing acceleration card (baseband processing module) and a clock processing module card may be used in hardware.

For another example, Mobile Edge Computing (MEC), whose architecture is shown in Figure 1e and Figure 1f, uses COTS general server (motherboard), artificial intelligence acceleration card (artificial intelligence acceleration module) and audio and video processing module card (audio and video) on the hardware. processing module).

Another example is a miniaturized mobile core network device, whose architecture is shown in Figure 1g and Figure 1h, and is composed of a COTS general server (mainboard), a network resource module and a storage resource module in hardware.

It can be seen that the existing wireless access network base stations, mobile edge computing equipment and wireless miniaturized core network products all have their own device forms and are of various types. This leads to complicated maintenance.

Application content

In view of this, the embodiment of the present application provides a fusion processing device to solve the problem of the wireless access network base station, mobile edge computing device, and wireless miniaturized core network product, etc. having their own device forms.

To achieve the above purpose, the embodiments of the present application provide the following technical solutions:

A fusion processing device, comprising: a backplane, a network resource module, a general computing resource module and a dedicated resource module on the backplane; wherein:

the backplane includes a first location area, a second location area and a third location area;

the network resource module is installed in the first location area;

the general-purpose computing resource module is installed in the second location area;

the dedicated resource module is installed in the third location area;

The network resource module, general computing resource module and special resource module on the backplane are all provided with communication interfaces;

The network resource modules, the general computing resource modules and the dedicated resource modules are interconnected through the communication interface and the bus on the backplane.

Optionally, the network resource module includes multiple network resource boards; the general-purpose computing resource module includes multiple general-purpose computing resource boards; the dedicated resource module includes multiple dedicated resource boards; the network resource board, the The general computing resource board and the dedicated resource board are provided with communication interfaces.

Optionally, the network resource module includes two network resource boards; the communication interfaces set on the network resource board, the general computing resource board and the dedicated resource board all include a business data interface and a control management interface; two network resource boards The bus used between the service data interface of the network resource board and the service data interface of the general computing resource board is: a double star bus; the bus used between the service data interface of the two network resource boards and the service data interface of the dedicated resource board are: double star bus; the bus used between the control management interface of the two network resource boards and the control management interface of the general computing resource board is: double star bus; the control management interface of the two network resource boards and the dedicated The bus used between the control and management interfaces of the resource board is: double star bus.

Optionally, the bus used for transmitting the clock signal between the two network resource boards and the dedicated resource board is: a double star bus; the bus used for transmitting the clock signal between the two network resource boards and the general computing resource board is: The bus is: double star bus.

Optionally, a point-to-point bus connection is used between the general-purpose computing resource board and the dedicated resource board.

Optionally, the dedicated resource boards are connected through a bus interface and a bidirectional ring bus; the bus interface is a CPRI/eCPRI interface, or a private interface; the network resource boards are connected through an Ethernet interface and a point-to-point bus. connect.

Optionally, the universal computing resource boards are connected point-to-point through an Ethernet interface and a bus.

Optionally, the types of chips on the multiple network resource boards include: at least one of a network multi-core processor chip, a switch chip, and a clock chip.

Optionally, the general-purpose computing resource board includes a general-purpose processor.

Optionally, the types of the multiple dedicated resource boards include: dedicated wireless baseband processing boards, dedicated high-reliability storage disk array boards, general-purpose FPGA large-scale logic boards, GPU/NPU processor boards, AI acceleration processing boards, and , at least one of the audio and video decoding and transcoding boards.

It can be seen that, in this embodiment of the present application, the backplane includes a first location area, a second location area, and a third location area, and a network resource module, a general-purpose computer module, and a dedicated resource module can be installed respectively. The network resource modules, the general computing resource modules and the dedicated resource modules are interconnected through the communication interface and the bus on the backplane. The above-mentioned network resource modules, general computing resource modules and dedicated resource modules provide the hardware architecture of products such as wireless access network base stations, mobile edge computing equipment, and wireless miniaturized core networks. Different products can use fusion processing equipment as hardware support , unified equipment form, maintenance is relatively simple. At the same time, the modular structure also facilitates the independent configuration and independent upgrade of each module.

Description of drawings

Fig. 1a is the physical structure of traditional wireless access network BBU baseband processing;

Figure 1b shows the principle architecture of BBU baseband processing in a traditional wireless access network;

Figure 1c is an exemplary physical structure of an existing distributed small cell DU base station;

Fig. 1d is a schematic diagram of the connection relationship of each component in the existing distributed small cell DU base station;

Figure 1e is an exemplary physical structure of an existing MEC;

Figure 1f is a schematic diagram of the connection relationship of each component in the existing MEC;

Fig. 1g is the exemplary physical structure of the existing miniaturized mobile core network equipment;

Fig. 1h is a schematic diagram of the connection relationship of each component in the existing miniaturized mobile core network equipment;

FIG. 2 is an overall structure of a fusion processing device provided by an embodiment of the present application;

FIG. 3 is an exemplary structure of a radio access network baseband product configured by a fusion processing device according to an embodiment of the application;

FIG. 4 is an exemplary physical structure corresponding to the radio access network baseband product shown in FIG. 3;

FIG. 5 is an exemplary physical structure corresponding to the radio access network baseband product shown in FIG. 3;

FIG. 6 is an exemplary structure of a mobile edge computing product configured by a fusion processing device according to an embodiment of the present application;

FIG. 7 is an exemplary physical structure corresponding to the mobile edge computing product shown in FIG. 6;

FIG. 8 is an exemplary physical structure corresponding to the mobile edge computing product shown in FIG. 6;

FIG. 9 is an exemplary structure of a miniaturized core network product configured by a fusion processing device according to an embodiment of the present application;

FIG. 10 is an exemplary physical structure corresponding to the miniaturized core network product shown in FIG. 9;

FIG. 11 is an exemplary physical structure corresponding to the miniaturized core network product shown in FIG. 9;

FIG. 12 is an exemplary structure of a fusion product provided by an embodiment of the application;

FIG. 13 is an exemplary physical structure corresponding to the fusion product shown in FIG. 12 .

Detailed ways

For the sake of reference and clarity, technical terms, abbreviations or abbreviations used hereinafter are summarized as follows:

MEC: Mobile Edge Computer, a mobile edge computing device that can be deployed between the wireless access network and the mobile core network. It is an edge computing (MEC) server built on a general hardware platform and artificial intelligence acceleration modules. It provides IT services at the edge of the mobile network. Service environment and cloud computing capabilities to reduce latency in network operations and service delivery;

BBU: Baseband Unit, baseband processing unit;

CN: Core Network Unit, core network unit;

ROC: RAID On Chip, RAID on chip;

COTS: Commercial off-the-shelf, using off-the-shelf products, purchased interface software or hardware products with open standard definitions, can save cost and time;

CT: Communication Technology, communication technology;

IT: Information Technology, information technology;

DT: Data Technology, data technology;

GNSS: Global Navigation Satellite System, global navigation satellite system;

CPRI: Common Public Radio Interface, common public radio interface;

eCPRI: enhanced Common Public Radio Interface, enhanced common public radio interface;

FPGA: Field Programmable Gate Array, field programmable gate array;

ASIC: Application Specific Integrated Circuit, application-specific integrated circuit;

GPU: Graphics Processing Unit, graphics processor;

NPU: Neural Network Processing Unit, neural network processor;

SATA: Serial Advanced Technology Attachment, Serial Advanced Technology Attachment, an industry-standard serial hardware drive interface.

The present application provides fusion processing equipment to solve the problems of wireless access network base stations, mobile edge computing equipment, and wireless miniaturized core network products that have their own equipment forms.

Figure 2 shows the overall structure of the above-mentioned fusion processing equipment, including:

Backplane 1, network resource module 2, general computing resource module 3 and dedicated resource module 4 on the backplane 1. In addition, a power supply module and a fan module can also be arranged on the backplane 1 .

in:

The backplane 1 includes a first location area, a second location area and a third location area;

The network resource module 2 is installed in the first location area;

The general computing resource module 3 is installed in the second location area;

The dedicated resource module 4 is installed in the third location area;

That is, each module has a corresponding installation position on the backplane.

The network resource module 2, the general computing resource module 3 and the dedicated resource module 4 on the backplane 1 are all provided with communication interfaces;

The network resource module 2 , the general computing resource module 3 and the dedicated resource module 4 are interconnected through the communication interface and the bus on the backplane 1 .

Taking the network resource module 2 and the general-purpose computing resource module 3 as examples, please refer to FIG. 2 . The part of the dotted line in FIG. 2 is the bus on the backplane 1 .

It should be noted that, in FIG. 2 , the connection between the network resource module 2 and the dedicated resource module 4 is not shown, and the two are actually interconnected through the communication interface and the bus in the backplane 1 . In addition, the communication interface is also not shown in FIG. 2 .

Based on the business characteristics of different products such as large bandwidth, low latency and multi-user, different buses (ETH/PCIE/RINGBUS interfaces, etc.) and communication interfaces can be selected for business interconnection.

In order to support the backup function, the network resource module 2 includes a plurality of network resource boards, and similarly, the general-purpose computing resource module 3 includes a plurality of general-purpose computing resource boards (that is, the computing resource boards in FIG. 2 ); the dedicated resource module 4 includes a plurality of Dedicated resource board.

Communication interfaces are arranged on the network resource board, the general computing resource board and the special resource board.

Taking network resource module 2 as an example, in the case of including N+1 (N greater than or equal to 1) network resource boards, 1+N inter-board backup can be implemented. If one of the network resource boards fails, services can be directly transferred to another. A single network resource board, therefore, a single point of failure has no impact on the business.

At the same time, the same network resource board, computing resource board or dedicated resource processing board can realize the load balancing function of processing capacity.

Each board is described in detail below.

1. Network resource board

The main functions of the network resource board are as follows:

Traffic classification of access service data packets, access control list, load balancing function;

GNSS\PTP\NTP high-precision clock synchronization, timekeeping function and various clock distribution functions;

Packet-based MAC address Layer 2 switching and IP address Layer 3 switching functions;

Shelf management, device management and operation management and maintenance functions.

The types of chips on the network resource board include, but are not limited to, one or more of network multi-core processor chips, switch chips, clock chips (eg, GNSS clock chips, PTP clock chips), and the like.

Among them, the network multi-core processing chip is used to complete data service processing, the switching chip can be used to realize service interaction between boards, the GNSS clock chip is used for GPS clock processing, and the PTP clock chip is used for PTP/1588 clock processing.

Multiple or one chips can be configured on the network resource board according to actual product needs, which will not be repeated here.

Second, the general computing resource board

The general purpose computing resource board includes a general purpose processor.

Wherein, the general-purpose processor exemplarily includes, but is not limited to, an X86 processor and an ARM processor.

The general computing resource board can flexibly realize baseband processing of wireless access network base stations, multi-service artificial intelligence processing of mobile edge computing, and miniaturized core network service processing. On the machine or Docker container, with an open software architecture, it is convenient for users to carry out secondary software development.

3. Dedicated resource board

The types of multiple dedicated resource boards include: dedicated wireless baseband processing boards, dedicated high-reliability storage disk array (ROC) boards, general-purpose FPGA large-scale logic boards, GPU/NPU processor boards, AI acceleration processing boards, and, audio and video Decode at least one of the transcoding boards.

Those skilled in the art can configure one or more types of dedicated resource boards according to product requirements.

Dedicated resource boards of the same type can communicate with each other.

As mentioned above, the modules are interconnected through the backplane bus. According to the characteristic requirements of different products, different high-speed interfaces (communication interfaces) and bus interconnection modes can be used, so that data interaction can meet the performance requirements of bandwidth and delay.

The following examples illustrate the types of high-speed interfaces and bus interconnection modes between boards:

1), the interconnection between the network resource board and other boards:

The communication interfaces set on the network resource board, the general computing resource board and the dedicated resource board can all include a service data interface and a control management interface.

Taking the network resource module including two network resource boards (that is, supporting the 1+1 backup function) as an example, between the service data interfaces of the above two network resource boards and the service data interfaces of the general computing resource board, the two network resource boards between the business data interface of the two network resource boards and the business data interface of the dedicated resource board, between the control and management interfaces of the two network resource boards and the control and management interface of the general computing resource board, and between the control and management interfaces of the two network resource boards and the dedicated resources The buses used between the control and management interfaces of the boards can all be: double star bus.

The service data interface and the control and management interface may specifically be a 10G/25G Ethernet interface.

Specifically, the Ethernet interface on the network resource board is interconnected with the switching chip on the network resource board, and a single channel can reach a bandwidth of 25 Gbit/s.

As mentioned above, the types of chips on the network resource board include clock chips, which generate clock signals. Then the bus used to transmit the clock signal between the two network resource boards and the dedicated resource board is: double star bus. Correspondingly, the bus used to transmit the clock signal between the two network resource boards and the general computing resource board is also a double star bus.

It should be noted that, since the clock signal is a pulse signal, no interface is required, and it can be directly connected by a double star bus.

2), the interconnection between the general computing resource board and the dedicated resource board:

Since there is a one-to-one relationship between the general-purpose computing resource board and the special-purpose resource board, a point-to-point bus connection is used between the general-purpose computing resource board and the special-purpose resource board.

The bus types include but are not limited to: PCIE bus with multi-channel bus, 10G/25G Ethernet bus, and SATA bus.

The multi-channel PCIE bus refers to: the PCIE bus can support multiple channels, such as 1 LANE (channel), 4 channels, 8 channels, or 16 channels.

Those skilled in the art select different interfaces based on the type of the dedicated resource board to meet the data transmission requirements of large bandwidth and low latency.

3), the interconnection between similar boards:

The network resource boards can be connected point-to-point through the Ethernet interface and the bus to support the real-time data synchronization function between the active and standby network resource boards.

Common Ethernet interfaces and bus point-to-point connections can be used between general computing resource boards to support real-time synchronization of business data and heartbeat detection. In addition, the general computing resource boards are directly connected through point-to-point Ethernet, which has good real-time performance and reliability, and does not depend on the connection of other resource boards.

The dedicated resource boards can be connected through a bus interface and a bidirectional ring bus. Specifically, the bus interface may be a CPRI/eCPRI interface, or a private interface, to support a service load sharing function between dedicated resource boards or a board-level fault self-healing function.

It can be seen that, in the embodiment of the present application, the backplane includes a first location area, a second location area, and a third location area, and a network resource module, a general-purpose computer module, and a dedicated resource module can be installed respectively. The network resource modules, the general computing resource modules and the dedicated resource modules are interconnected through the communication interface and the bus on the backplane. The above-mentioned network resource module, general computing resource module, and dedicated resource module provide hardware architectures for products such as wireless access network base stations, mobile edge computing devices, and wireless miniaturized core networks.

Different products can use fusion processing equipment as hardware support, which unifies the equipment form and is relatively simple to maintain. At the same time, the modular structure also facilitates the independent configuration and independent upgrade of each module.

The following takes a specific product as an example to introduce how to configure each module of the above-mentioned fusion processing device.

1. Wireless access network baseband products:

Please refer to FIG. 3 to FIG. 5 (FIG. 4 and FIG. 5 are diagrams of the same hardware architecture in different directions), the hardware architecture of the radio access network baseband product exemplarily includes:

1. General computing resource board (general computing board):

X86 processor and ARM v8 processor can be set on the general computing resource board.

Computing units can be designed based on the above-mentioned X86 or ARMv8 processors to support virtualization and cloudification of radio access network baseband products.

2. Dedicated wireless baseband processing board (baseband processing board):

In this embodiment, the dedicated resource board specifically includes a baseband processing board.

Specifically, the baseband processing board may include a dedicated baseband SOC, an FPGA or a dedicated ASIC chip.

2G, 4G, and 5G physical layer processing functions can be designed based on the above-mentioned dedicated baseband SOC, FPGA or dedicated ASIC chip to meet the requirements of multi-standard base stations.

3. Network resource board:

The chips on the network resource board include: a network multi-core processor chip, a switch chip, and a clock chip (eg, a GNSS clock chip, a PTP clock chip). It can provide multiple 10G, 25G and other network interfaces to access business traffic (implemented by network multi-core processor chips or network switching chips), realize service switching functions (implemented by switching chips), GNSS and PTP clock processing and distribution Function (implemented by GNSS clock chip, PTP clock chip), etc.

In addition, it can also include fan boards, active and standby power supply boards, etc.

Second, mobile edge computing products:

Referring to Figures 6-8 (Figures 7 and 8 are diagrams of the same hardware architecture in different directions), the hardware architecture of mobile edge computing products exemplarily includes:

1. General computing resource board (general computing board):

2. AI acceleration processing board and audio and video processing board:

In this embodiment, the dedicated resource board specifically includes an AI acceleration processing board and an audio and video processing board. Among them, the AI acceleration processing board can accelerate artificial intelligence inference and training based on GPU, AI-specific ASIC chip or FPGA; the audio and video processing board can perform audio and video decoding or transcoding processing, mainly to accelerate audio and video processing functions.

The bus between the computing resource board and the AI acceleration processing board/audio and video processing board is the bus on the backplane.

10G-KR in Figure 6 represents an Ethernet interface, and 12V represents the power supply voltage for each board.

3. Network resource board:

In this embodiment, the network resource board can provide multiple 10G, 25G and other network interfaces with various rates to access service traffic, and perform preprocessing on the service traffic.

Three, miniaturized core network products:

Referring to Figures 9-11 (Figures 10 and 11 are diagrams of the same hardware architecture in different directions), the hardware architecture of the miniaturized core network product exemplarily includes:

1. General computing resource board (general computing board):

2. Network resource board:

3. Storage resource board:

In this embodiment, the dedicated resource board specifically includes a storage resource board. The storage resource board is based on the RAID function of the SAS controller, which can realize high reliability and large-capacity HDD or SDD storage system, and a single machine can support 12 HDD or SDD hard disks.

Fourth, the radio access network baseband base station, mobile edge computing and miniaturized core network integration products:

Please refer to Figure 12-13, the converged product of the radio access network baseband BBU, mobile edge computing and miniaturized core network products, the exemplary hardware architecture includes:

1. General computing resource board (general computing board):

Based on virtual machine VM, container Docker technology and software-defined product features, RAN, MEC, and NGC network functions (NFV) can run on one or more general-purpose computing boards to form an edge cloud open business system.

2. Network resource board:

3. AI acceleration processing board:

AI acceleration processing boards can accelerate artificial intelligence inference and training based on GPUs, AI-specific ASIC chips or FPGAs.

4. Storage resource board:

The storage resource board is based on the RAID function of the SAS controller, which can realize high reliability and large-capacity HDD or SDD storage system, and a single machine can support 12 HDD or SDD hard disks.

5. Dedicated wireless baseband processing board (baseband processing board):

That is, in this embodiment, the dedicated resource board specifically includes an AI acceleration processing board, a storage resource board, and a baseband processing board.

With the rapid deployment of 5G technology in all walks of life, industry 5G private networks and industry applications are becoming more and more extensive, including the application of 5G in rail transportation, 5G networks combined with edge computing MEC + artificial intelligence AI and blockchain technology applications, 5G For applications such as smart parks and smart factories, it is hoped that 5G base station systems, edge computing equipment and core network systems can be effectively integrated in hardware, miniaturized equipment, and virtualized technology for integrated deployment of business software.

The fusion processing equipment claimed in this application adopts a hardware modular design, and adopts a personalized bus interconnection mode and interface type between module resources, so as to meet the requirements of the fusion wireless access network base station, edge computing and wireless core network equipment.

To sum up, the technical solution provided by this application has the following advantages:

1. Separation of general-purpose processing hardware (X86 or ARMv8) and dedicated processing hardware (FPGA/GPU/ASIC) to facilitate independent configuration and independent upgrade of each module;

2. Independent computing resources, network resources and storage resources can be flexibly configured according to different product requirements;

3. The device form (the device form is hardware composition on the one hand, and the functional characteristics of the product on the other hand) can be realized through modular configuration and software-defined function (SDN);

4. Flexibly carry functional equipment such as wireless base station communication (RAN), edge computing (MEC) and core network (NGC) to realize seamless switching of equipment and mutual resource opening;

5. Comprehensive CT-based hardware and software (such as wireless multi-mode base stations), IT-based (such as miniaturized core network equipment) and DT-based (edge computing equipment), supporting in-device resource pooling, function virtualization and edge board-level cloud change.

6. Using fine-grained modular equipment technology and multi-modular design, each module can be distributed for power supply and multi-point heat dissipation, effectively solving the problems of single-point heat dissipation and single-point power supply. However, if multiple modules are designed into one board as in the prior art, and one board has multiple chips with high power consumption to dissipate heat, the problem of heat dissipation is difficult to solve.

The description of the disclosed embodiments enables any person skilled in the art to make or use the application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the generic principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, this application is not intended to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

A fusion processing device, comprising: a backplane, a network resource module, a general computing resource module and a dedicated resource module on the backplane; wherein:

the backplane includes a first location area, a second location area and a third location area;

the network resource module is installed in the first location area;

the general-purpose computing resource module is installed in the second location area;

the dedicated resource module is installed in the third location area;

The network resource module, general computing resource module and special resource module on the backplane are all provided with communication interfaces;

The network resource modules, the general computing resource modules and the dedicated resource modules are interconnected through the communication interface and the bus on the backplane.
The apparatus of claim 1, wherein:

The network resource module includes a plurality of network resource boards;

The general-purpose computing resource module includes a plurality of general-purpose computing resource boards;

The dedicated resource module includes multiple dedicated resource boards;

Communication interfaces are provided on the network resource board, the general computing resource board and the dedicated resource board.
The apparatus of claim 1, wherein:

The network resource module includes two network resource boards;

The communication interfaces set on the network resource board, the general computing resource board and the special resource board all include a business data interface and a control management interface;

The bus used between the service data interface of the two network resource boards and the service data interface of the general computing resource board is: a double star bus;

The bus used between the service data interface of the two network resource boards and the service data interface of the dedicated resource board is: double star bus;

The bus used between the control and management interfaces of the two network resource boards and the control and management interface of the universal computing resource board is: a double star bus;

The bus used between the control and management interfaces of the two network resource boards and the control and management interface of the dedicated resource board is a double star bus.
The apparatus of claim 3, wherein:

The bus used to transmit the clock signal between the two network resource boards and the dedicated resource board is: double star bus;

The bus used to transmit the clock signal between the two network resource boards and the general computing resource board is a double star bus.
The apparatus of claim 2, wherein:

A point-to-point bus connection is used between the general-purpose computing resource board and the dedicated resource board.
The apparatus of claim 2, wherein:

The dedicated resource boards are connected through a bus interface and a bidirectional ring bus; the bus interface is a CPRI/eCPRI interface, or a private interface;

The network resource boards are connected point-to-point through an Ethernet interface and a bus.
The apparatus of claim 2, wherein:

The universal computing resource boards are connected point-to-point through an Ethernet interface and a bus.
The apparatus of claim 2, wherein:

The types of chips on the multiple network resource boards include: at least one of a network multi-core processor chip, a switch chip, and a clock chip.
The apparatus of claim 2, wherein:

The general-purpose computing resource board includes a general-purpose processor.
The apparatus of claim 2, wherein:

The types of the multiple dedicated resource boards include: dedicated wireless baseband processing boards, dedicated high-reliability storage disk array boards, general-purpose FPGA large-scale logic boards, GPU/NPU processor boards, AI acceleration processing boards, and audio and video decoding. At least one of the transcoding boards.